CN120908582A - Lightning arrester fault indication judging method and indicating device - Google Patents
Lightning arrester fault indication judging method and indicating deviceInfo
- Publication number
- CN120908582A CN120908582A CN202511438371.2A CN202511438371A CN120908582A CN 120908582 A CN120908582 A CN 120908582A CN 202511438371 A CN202511438371 A CN 202511438371A CN 120908582 A CN120908582 A CN 120908582A
- Authority
- CN
- China
- Prior art keywords
- lightning arrester
- data
- lightning
- leakage current
- characteristic
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention relates to the technical field of lightning arrester fault detection, in particular to a lightning arrester fault indication judging method and an indicating device. A fault indication judging method of a lightning arrester comprises the steps of S1, historical data acquisition, S2, feature vector generation, S3, a neural network with a multi-head attention mechanism, S4, model training and optimization, and S5, real-time monitoring. According to the lightning arrester fault judging method, the lightning strike current data, the leakage current data and the temperature difference data between the lightning arrester and the environment are collected, so that a neural network model which is input in multiple channels and has a multi-head attention judging mechanism is built, when lightning strike occurs to the lightning arrester, the model reacts at the first time and reflects the fault condition to the indicating device, and for aging condition detection with long time span, accurate detection of aging influence factors is achieved, and fault indication to the lightning arrester can be achieved through the lightning arrester fault judging method built by the model.
Description
Technical Field
The invention relates to the technical field of lightning arrester fault detection, in particular to a lightning arrester fault indication judging method and an indicating device.
Background
The gapless zinc oxide arrester is widely used in a power system, plays an important role in maintaining the stability of a power transmission and distribution line, and the conventional arrester detection method is to apply direct current and detect leakage current, namely, a detector applies direct current to the arrester outside, and the lightning protection performance of the current arrester is detected by detecting the magnitude of the leakage current inside the arrester. However, for the live lightning arrester actually installed in the transmission line, the live lightning arrester works in dangerous alternating current environment, and it is difficult for a detector to apply direct current voltage and current in the working process, so that the difficulty of detecting whether the live lightning arrester is faulty is greatly increased, and therefore, the method capable of achieving fault indication judgment through sensor data of the lightning arrester is considered to be developed.
Disclosure of Invention
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and the appended drawings.
The invention aims to overcome the defects, and provides a lightning arrester fault indication judging method and an indicating device, wherein a neural network model which is input in multiple channels and has a multi-head attention judging mechanism is built by collecting lightning stroke current data, leakage current data and temperature difference data between the lightning arrester and the environment, the lightning stroke current data is input into a multi-head attention module after lightning stroke current characteristics are extracted through a one-dimensional convolution neural network, the attention module is provided with the highest weight, therefore, when the lightning arrester has an emergency of lightning stroke, the model reacts at the first time and reflects the fault condition to the indicating device, and for aging condition detection with a longer time span, the lightning arrester fault judging method built through the model can be realized by inputting the leakage current data and the temperature difference data between the lightning arrester and the environment into an LSTM neural network in combination and setting the weights for the data conditions of the white days and the night.
The invention provides a fault indication judging method of a lightning arrester, which comprises the following steps:
s1, historical data acquisition, namely acquiring lightning stroke current data by a lightning stroke sensor in an existing lightning arrester, acquiring leakage current data by a leakage current sensor, arranging a photovoltaic voltage sensor at the top of the lightning arrester, acquiring photovoltaic voltage data, acquiring the temperature of the lightning arrester by a temperature sensor of the lightning arrester, and acquiring the external environment temperature by an environment temperature sensor;
S2, generating a feature vector:
S21, generating a lightning current sequence for the collected lightning current data, inputting the lightning current sequence into a one-dimensional convolutional neural network, and generating lightning current characteristic data;
S22, judging whether the current weather is daytime or night according to the photovoltaic voltage data, and forming a gating value;
s23, inputting acquired all-day leakage current data into an LSTM1 neural network, and respectively generating leakage current characteristic data with daytime and night information;
S24, inputting all-day temperature difference data formed by the temperature of the lightning arrester and the ambient temperature into an LSTM2 neural network, and respectively generating temperature difference characteristic data with daytime and night information;
s25, fusing the weighted gate control value with leakage current characteristic data to form one-dimensional final leakage current characteristic, fusing the weighted gate control value with temperature difference characteristic data to form one-dimensional final temperature difference characteristic, and forming aging influence characteristic data by the final leakage current characteristic and the final temperature difference characteristic through a characteristic superposition algorithm;
S3, adding a multi-head attention mechanism to the lightning current characteristic data and the aging influence characteristic data generated in the step S2, setting different parts in a plurality of attention head attention sequences and generating corresponding weights, carrying out weighted summation on the outputs of the plurality of heads, and obtaining diagnosis results of the lightning arrester after fusion classification, wherein the diagnosis results comprise normal, lightning stroke and aging;
S4, training and optimizing the model, namely training the model established in the step S1-S3 by setting a test set and a verification set to form a mature fault diagnosis model of the lightning arrester;
s5, real-time monitoring, namely inputting the data of the lightning arrester into a fault diagnosis model of the lightning arrester to be monitored in real time, and outputting a diagnosis result to realize real-time monitoring.
In some embodiments, in step S21, the one-dimensional convolutional neural network specifically includes an input layer, a convolutional layer, an activation function, and a pooling layer, where the input layer of the one-dimensional convolutional neural network is configured to receive lightning current data to generate a lightning current sequence, the convolutional layer is a one-dimensional convolutional, the activation function is a ReLU, the pooling layer adopts one-dimensional pooling with global average, and the pooling layer outputs lightning current feature data to be input to the multi-head attention module.
In some embodiments, in step S22, the collected photovoltaic voltage data is used as a mark for distinguishing day and night, and the influence of short-time clouding shielding on the photovoltaic voltage data is filtered, when the photovoltaic voltage is less than 0.3V and lasts for 1h, the night is considered, when the photovoltaic voltage is greater than or equal to 0.3V and lasts for 1h, the day is considered, a sigmoid gating function is set according to the judging condition, and a gating value g between 0 and 1 is generated.
In some embodiments, in step S25, the specific step of forming the aging-influencing feature data by using the feature stacking algorithm is to align the one-dimensional final leakage current feature and the final temperature difference feature, and add the data at the corresponding positions of the two features to form new aging-influencing feature data, and calculate the duty ratio of the final leakage current feature and the final temperature difference feature in the new aging-influencing feature.
In some embodiments, for the duty ratio of the final leakage current feature and the final temperature difference feature in the new aging influence feature number, if the diagnosis result finally output by the lightning arrester fault diagnosis model is aging, the duty ratio result is displayed together with the aging feature, and if the output result is normal or lightning stroke, the duty ratio result is not displayed.
In some embodiments, in step S3, the specific step of setting the neural network with a multi-headed attention mechanism includes:
S41, splicing lightning stroke current characteristic data and aging influence characteristic data to serve as input of a multi-head attention module;
s42, dividing an input sequence into 3 heads, and adding position codes;
S43, performing three different linear transformations on the input sequence added with the position codes to obtain a query Q, a key K and a value V;
S44, normalizing the attention score by using a Softmax function to obtain the attention weight;
s45, carrying out weighted summation on the attention weight and the value V;
s46, fusing the characteristics through the full connection layer, outputting through the linear layer, and outputting a diagnosis result of the lightning arrester.
In some embodiments, in the training and optimizing process of step S4, the mean square error is used as a loss function, training and optimization is implemented through an optimizer, and an Adam optimizer is selected as an optimization method of model parameters.
A lightning arrester fault indication device comprising:
a housing mounted on the arrester;
The control center is internally provided with a lightning arrester fault diagnosis model, and is arranged in the shell;
The display screen is connected with the control center and is arranged on the surface of the shell;
an indication fan provided with three blades and having different colors;
The indicating fan is electrically connected with the control center, and rotates to a color corresponding to the diagnosis result when the arrester fault diagnosis model outputs the diagnosis result.
By adopting the technical scheme, the invention has the beneficial effects that:
According to the lightning arrester fault detection method, the lightning strike current data, the leakage current data and the temperature difference data between the lightning arrester and the environment are collected, so that a multi-channel input neural network model with a multi-head attention judgment mechanism is built, the set lightning strike current data is input into the multi-head attention module after lightning strike current characteristics are extracted through the one-dimensional convolution neural network, the attention module is provided with the highest weight, therefore, when lightning strike emergency occurs in the lightning arrester, the model reacts at the first time and reflects the fault condition to the indicating device, for aging condition detection with a longer time span, the lightning arrester fault condition is input into the LSTM neural network through the combination of the leakage current data and the temperature difference data between the lightning arrester and the environment, and the accurate detection of aging influence factors is achieved, and fault indication of the lightning arrester can be achieved through the lightning arrester fault judgment method built through the model.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
It is apparent that these and other objects of the present invention will become more apparent from the following detailed description of the preferred embodiments, which is to be read in connection with the accompanying drawings and figures.
The foregoing and other objects, features and advantages of the invention will be apparent from the following more particular description of the preferred embodiments, as illustrated in the accompanying drawings.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate the invention and together with the embodiments of the invention, serve to explain the invention, without limitation to the invention.
In the drawings, like parts are designated with like reference numerals and are illustrated schematically and are not necessarily drawn to scale.
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only one or several embodiments of the invention, and that other drawings can be obtained according to such drawings without inventive effort to a person skilled in the art.
FIG. 1 is a schematic overall flow chart of a fault indication determination method according to some embodiments of the present invention;
FIG. 2 is a schematic diagram of the overall architecture of a multi-channel input-to-multi-head attention mechanism neural network according to some embodiments of the present invention;
FIG. 3 is a schematic diagram of a process for extracting timing characteristics of photovoltaic voltages, temperature difference sequences, and leakage currents through an LSTM neural network in accordance with some embodiments of the present invention;
FIG. 4 is a schematic diagram of an LSTM neural network in accordance with some embodiments of the invention;
FIG. 5 is a schematic diagram of a process of weighting and superimposing temperature difference characteristics and leakage current characteristics in some embodiments of the invention.
Fig. 6 is a schematic view of an internal structure of a fault indication device for an arrester according to some embodiments of the present invention.
The main reference numerals illustrate:
1. the fan comprises a shell, a control center and an indication fan.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the following detailed description. It should be understood that the detailed description is presented merely to illustrate the invention, and is not intended to limit the invention.
In addition, in the description of the present invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the drawings, are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed, directly connected, indirectly connected via an intermediate medium, or in communication with each other between two elements or in an interaction relationship between two elements. However, it is noted that direct connection indicates that the two bodies connected together do not form a connection relationship through a transition structure, but are connected together to form a whole through a connection structure. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
In the present invention, unless expressly stated or limited otherwise, a first feature "up" or "down" a second feature may be the first and second features in direct contact, or the first and second features in indirect contact via an intervening medium. In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Referring to fig. 1-2, fig. 1 is a schematic overall flow chart of a fault indication judging method according to some embodiments of the present invention, and fig. 2 is a schematic overall structure diagram of a multi-channel input-to-multi-head attention mechanism neural network according to some embodiments of the present invention.
According to some embodiments of the present invention, there is provided a lightning arrester fault indication judging method, including:
s1, historical data acquisition, namely acquiring lightning stroke current data by a lightning stroke sensor in an existing lightning arrester, acquiring leakage current data by a leakage current sensor, arranging a photovoltaic voltage sensor at the top of the lightning arrester, acquiring photovoltaic voltage data, acquiring the temperature of the lightning arrester by a temperature sensor of the lightning arrester, and acquiring the external environment temperature by an environment temperature sensor;
The lightning stroke sensor is used for collecting lightning stroke current data, when the lightning stroke current data is suddenly increased, the lightning arrester is currently subjected to lightning stroke, the zinc oxide lightning arrester in a conventional state can bear the impact of 20-30 times of nominal discharge current, but if the single current exceeds the rated value, the lightning arrester is possibly broken down and short-circuited, and the lightning arrester is reminded of timely replacement of the lightning arrester by jumping to the breakdown state of the lightning arrester;
The leakage current data are used for detecting the aging state of the lightning arrester, the lightning arrester can be aged with time after bearing the impact of the nominal discharge current every time according to the lightning stroke detection process, when the leakage current is increased to a certain value, the lightning arrester cannot be used again, a maintenance person needs to be reminded of timely replacing the lightning arrester, the leakage current data of the lightning arrester can be gradually increased with time, the leakage current can be changed along with the temperature difference, when the temperature difference between the ambient temperature detected by the photovoltaic voltage sensor and the temperature of the lightning arrester collected by the temperature sensor is large, the leakage current can be correspondingly increased, when the temperature difference between the ambient temperature and the temperature of the lightning arrester is small, the leakage current can be correspondingly reduced, and therefore, when the lightning arrester is required to be considered to be aged, the lightning arrester is aged with time, the fault condition occurs, and the fault condition occurs due to overlarge temperature difference.
Referring to fig. 3-4, fig. 3 is a schematic diagram illustrating a process of extracting time sequence characteristics of photovoltaic voltage, temperature difference sequence and leakage current through an LSTM neural network according to some embodiments of the present invention, and fig. 4 is a schematic diagram illustrating a structure of the LSTM neural network according to some embodiments of the present invention.
S2, generating a feature vector:
S21, generating a lightning current sequence for the collected lightning current data, inputting the lightning current sequence into a one-dimensional convolutional neural network, and generating lightning current characteristic data;
The one-dimensional convolution neural network specifically comprises an input layer, a convolution layer, an activation function and a pooling layer, wherein the input layer of the one-dimensional convolution neural network is used for receiving lightning current data to generate a lightning current sequence, the convolution layer is one-dimensional convolution, the activation function is ReLU, the pooling layer adopts global average one-dimensional pooling, and the pooling layer outputs lightning current characteristic data and inputs the lightning current characteristic data into the multi-head attention module;
S22, judging whether the current weather is daytime or night according to the photovoltaic voltage data, and forming a gating value;
The collected photovoltaic voltage data is used as a mark for distinguishing daytime and night, meanwhile, the influence of short-time clouding shielding on the photovoltaic voltage data is filtered, the night is considered when the photovoltaic voltage is smaller than 0.3V and lasts for 1h, the daytime is considered when the photovoltaic voltage is larger than or equal to 0.3V and lasts for 1h, a sigmoid gating function is set according to a judging condition, and a gating value g between 0 and 1 is generated;
s23, inputting acquired all-day leakage current data into an LSTM1 neural network, and respectively generating leakage current characteristic data with daytime and night information;
S24, inputting all-day temperature difference data formed by the temperature of the lightning arrester and the ambient temperature into an LSTM2 neural network, and respectively generating temperature difference characteristic data with daytime and night information;
During the daytime, due to the influence of solar radiation, the temperature difference between the lightning arrester and the environment is larger, the temperature difference between the lightning arrester and the environment is collected, the temperature difference sequence is input into the LSTM1 neural network to form daytime temperature difference data characteristics, leakage current during daytime operation is measured, the temperature difference sequence is input into the LSTM2 neural network to form daytime leakage current data characteristics, and the LSTM neural network learns to enable the leakage current data characteristics and the temperature difference data characteristics to have daytime time sequence information;
At night, as solar radiation is reduced, under a normal working state, the temperature difference between the lightning arrester and the environment is smaller, the temperature difference between the lightning arrester and the environment is collected, a temperature difference sequence is input into an LSTM1 neural network to form a night temperature difference data characteristic, leakage current under the night work is measured, the temperature difference sequence is input into an LSTM2 neural network to form a night leakage current data characteristic, and the LSTM neural network learns to enable the leakage current data characteristic and the temperature difference data characteristic to have timing information at night;
The reason why the time sequence characteristics of the day and the night are required to be studied separately is that in the day state, the leakage current is larger due to the temperature difference, and the same current is judged to be in a fault state by the system in the night state, so that the difference between the day and the night is required to be considered, and the system is attached to the actual use state of the lightning arrester;
As shown in FIG. 4, the specific step of capturing the time dependence of the leakage current characteristic data or the temperature difference characteristic data is to take the leakage current characteristic sequence or the temperature difference characteristic sequence as Input to the forgetting gate and input gate respectively, the forgetting gate discards useless memory to generate cell stateThe input gate screens new information to generate candidate cell state, and the cell state is updated by combining the result generated by the forgetting gate and the input gateThe active function of the hidden layer is a sigmoid function and a tanh function, and the unit number defaults to 128; in the hidden state of the previous time step, And outputting leakage current characteristic data or temperature difference characteristic data with time sequence information after passing through the full-connection layer to obtain the hidden state after updating the cells.
Referring to FIG. 5, FIG. 5 is a schematic diagram of a process of weighting and superimposing temperature difference characteristics and leakage current characteristics in some embodiments of the invention;
s25, fusing the weighted gate control value with leakage current characteristic data to form one-dimensional final leakage current characteristic, fusing the weighted gate control value with temperature difference characteristic data to form one-dimensional final temperature difference characteristic, and forming aging influence characteristic data by the final leakage current characteristic and the final temperature difference characteristic through a characteristic superposition algorithm;
The day and night characteristics of the temperature difference and the leakage current are respectively learned through the LSTM neural network, a large amount of data is needed, the accurate learning needs a long time, the speed of judging the day and night state of the model can be improved through a mode of manually adding prior conditions, namely, the gating value in the step S22 is introduced, the night is considered when the photovoltaic voltage is smaller than 0.3V and lasts for 1h, the day is considered when the photovoltaic voltage is larger than or equal to 0.3V and lasts for 1h, and a sigmoid gating function is set according to the judging conditions;
Taking the daytime temperature difference and the night temperature difference as examples, as a detector sets a definite threshold value for the photovoltaic voltage of the lightning arrester according to actual experience, the sigmoid gating function can be set to be very steep, so that g is very close to 0 (night) or 1 (day), the LSTM neural network firstly forms the day and night characteristics of preliminary temperature difference and leakage current, and finally the final temperature difference characteristics = g are the daytime temperature difference characteristics + (1-g) and the night temperature difference characteristics, so that the model does not need to accurately learn the day and night rules, becomes easier to train, and has quicker conclusion and better universality;
The specific step of forming aging influence characteristic data through a characteristic superposition algorithm is that one-dimensional final leakage current characteristic and final temperature difference characteristic are aligned, data on corresponding positions of the final leakage current characteristic and the final temperature difference characteristic are added to form new aging influence characteristic data, meanwhile, the duty ratio of the final leakage current characteristic and the final temperature difference characteristic in new aging influence characteristic number is calculated, and for the duty ratio of the final leakage current characteristic and the final temperature difference characteristic in the new aging influence characteristic number, if a diagnosis result finally output by a lightning arrester fault diagnosis model is aging, the duty ratio result is displayed together with the aging characteristic, and if the output result is normal or lightning stroke, the duty ratio result is not displayed;
In the application, the characteristic superposition algorithm can fuse the data of the temperature sensor and the leakage current sensor, so as to improve the decision capability and accuracy of a detection result, and simultaneously, the characteristic superposition algorithm can effectively process and analyze a large amount of data, has suitability for detecting aging data with longer time span and can improve the data processing capability of a model.
S3, adding a multi-head attention mechanism to the lightning current characteristic data and the aging influence characteristic data generated in the step S2, setting different parts in a plurality of attention head attention sequences and generating corresponding weights, carrying out weighted summation on the outputs of the plurality of heads, and obtaining diagnosis results of the lightning arrester after fusion classification, wherein the diagnosis results comprise normal, lightning stroke and aging;
the specific steps of setting the neural network with the multi-head attention mechanism include:
s41, splicing lightning stroke current characteristic data and aging influence characteristic data to be used as input of a multi-head attention module ;
When the one-dimensional splicing is used for processing multi-mode data or a plurality of feature sources, feature vectors of different sources can be combined together through the one-dimensional splicing to form a comprehensive feature representation, so that the neural network is helpful for capturing the diversity and complexity of the data better, and the dimension of input features can be increased through splicing a plurality of feature vectors, so that the expression capacity of a model is improved;
S42, inputting the sequence Dividing into 3 heads, and adding position codes;
Dimension of each head , wherein,Adding corresponding position codes into the input for the total dimension of the model;
wherein P is a position code, l is a time sequence position corresponding to the position code, and k is 0 to An integer value in between;
for sequence length L and model overall dimension Creating an LA position encoding matrix PE of (a);
the position-coding matrix PE is added to the input sequence: x= +PE such that the input data for each time stepThe method comprises the steps of including position information, wherein X is an input sequence added with position codes;
S43, performing three different linear transformations on the input sequence added with the position codes to obtain a query Q, a key K and a value V;
wherein, the 、、Respectively a weight matrix;
for each query Q, calculate its attention score with all keys K ;
S44, calculating the output of each attention head by using a Softmax function;
wherein, the For the i-th attention head,A value V for the ith attention header;
s45, fusing the characteristics through a full connection layer, outputting through three linear layers, and outputting a diagnosis result of the lightning arrester;
the final output of the fused multi-head attention mechanism is as follows:
Wherein head 1 to head 3 are the outputs of three attention heads respectively, For mapping the spliced vectors to a final output space,Is the final output of the multi-headed attention mechanism.
S4, training and optimizing the model, namely training the model established in the step S1-S3 by setting a test set and a verification set to form a mature fault diagnosis model of the lightning arrester;
In the training and optimizing process of the step S4, the mean square error is used as a loss function, training and optimizing are achieved through an optimizer, and an Adam optimizer is selected as an optimizing method of model parameters.
S5, real-time monitoring, namely inputting the data of the lightning arrester into a fault diagnosis model of the lightning arrester to be monitored in real time, and outputting a diagnosis result to realize real-time monitoring.
Referring to fig. 6, fig. 6 is a schematic view illustrating an internal structure of a fault indication device for an arrester according to some embodiments of the present invention.
According to some embodiments of the present invention, there is provided a lightning arrester fault indication device, including:
a housing 1 mounted on the lightning arrester;
A control center 2 having a lightning arrester fault diagnosis model disposed therein, the control center 2 being disposed in the housing 1;
The display screen is connected with the control center 2 and is arranged on the surface of the shell;
An indicating fan 3 provided with three blades and having different colors;
The indicating fan 3 is electrically connected with the control center 2, and when the arrester fault diagnosis model outputs a diagnosis result, the indicating fan 3 rotates to a color corresponding to the diagnosis result.
In the actual use process, lightning stroke faults are represented by red fan blades, ageing faults are represented by yellow fan blades, the ageing faults are represented by green fan blades in a normal state, a PCB circuit board integrated with a control center 2 is arranged above the indication fan 3 and shields the two fan blades, when the current fault problem is judged, the red or yellow fan blades are rotated to the lower side and can be observed by the outside, and if the lightning arrester works normally, the green fan blades are rotated to the lower side.
It is to be understood that the disclosed embodiments are not limited to the specific process steps or materials disclosed herein, but are intended to extend to equivalents of such features as would be understood by one of ordinary skill in the relevant arts. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.
Reference in the specification to "an embodiment" means that a particular feature, or characteristic, described in connection with the embodiment is included in at least one embodiment of the invention. Thus, appearances of the phrase or "an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment.
Furthermore, the described features or characteristics may be combined in any other suitable manner in one or more embodiments. In the above description, certain specific details are provided, such as thicknesses, numbers, etc., to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention can be practiced without one or more of the specific details, or with other methods, components, materials, etc.
Claims (8)
1. A fault indication judging method for a lightning arrester is characterized by comprising the following steps of
S1, historical data acquisition, namely acquiring lightning stroke current data by a lightning stroke sensor in an existing lightning arrester, acquiring leakage current data by a leakage current sensor, arranging a photovoltaic voltage sensor at the top of the lightning arrester, acquiring photovoltaic voltage data, acquiring the temperature of the lightning arrester by a temperature sensor of the lightning arrester, and acquiring the external environment temperature by an environment temperature sensor;
S2, generating a feature vector:
S21, generating a lightning current sequence for the collected lightning current data, inputting the lightning current sequence into a one-dimensional convolutional neural network, and generating lightning current characteristic data;
S22, judging whether the current weather is daytime or night according to the photovoltaic voltage data, and forming a gating value;
s23, inputting acquired all-day leakage current data into an LSTM1 neural network, and respectively generating leakage current characteristic data with daytime and night information;
S24, inputting all-day temperature difference data formed by the temperature of the lightning arrester and the ambient temperature into an LSTM2 neural network, and respectively generating temperature difference characteristic data with daytime and night information;
s25, fusing the weighted gate control value with leakage current characteristic data to form one-dimensional final leakage current characteristic, fusing the weighted gate control value with temperature difference characteristic data to form one-dimensional final temperature difference characteristic, and forming aging influence characteristic data by the final leakage current characteristic and the final temperature difference characteristic through a characteristic superposition algorithm;
S3, adding a multi-head attention mechanism to the lightning current characteristic data and the aging influence characteristic data generated in the step S2, setting different parts in a plurality of attention head attention sequences and generating corresponding weights, carrying out weighted summation on the outputs of the plurality of heads, and obtaining diagnosis results of the lightning arrester after fusion classification, wherein the diagnosis results comprise normal, lightning stroke and aging;
S4, training and optimizing the model, namely training the model established in the step S1-S3 by setting a test set and a verification set to form a mature fault diagnosis model of the lightning arrester;
s5, real-time monitoring, namely inputting the data of the lightning arrester into a fault diagnosis model of the lightning arrester to be monitored in real time, and outputting a diagnosis result to realize real-time monitoring.
2. The lightning arrester fault indication judging method according to claim 1, wherein in step S21, the one-dimensional convolutional neural network specifically includes an input layer, a convolutional layer, an activation function and a pooling layer, the input layer of the one-dimensional convolutional neural network is used for receiving lightning current data to generate a lightning current sequence, the convolutional layer is one-dimensional convolutional, the activation function is ReLU, the pooling layer adopts global average one-dimensional pooling, and the pooling layer outputs lightning current characteristic data to be input into the multi-head attention module.
3. The lightning arrester fault indication judging method according to claim 1, wherein in step S22, the collected photovoltaic voltage data is used as a mark for distinguishing day and night, and the influence of short-time clouding shielding on the photovoltaic voltage data is filtered, the night is considered when the photovoltaic voltage is less than 0.3V and lasts for 1h, the day is considered when the photovoltaic voltage is greater than or equal to 0.3V and lasts for 1h, a sigmoid gating function is set according to the judging condition, and a gating value g between 0 and 1 is generated.
4. The lightning arrester fault indication judging method according to claim 1, wherein in the step S25, the specific step of forming the aging-affecting feature data through the feature superposition algorithm is to align the one-dimensional final leakage current feature and the final temperature difference feature, add the data at the corresponding positions of the two features to form new aging-affecting feature data, and calculate the duty ratio of the final leakage current feature and the final temperature difference feature in the new aging-affecting feature number.
5. The method according to claim 4, wherein for the duty ratio of the final leakage current characteristic and the final temperature difference characteristic in the new aging influence characteristic number, if the diagnosis result finally output by the arrester fault diagnosis model is aging, the duty ratio result is displayed together with the aging characteristic, and if the output result is normal or lightning stroke, the duty ratio result is not displayed.
6. The lightning arrester fault indication judgment method according to claim 1, wherein in step S3, the specific step of setting a neural network having a multi-headed attention mechanism includes:
S41, splicing lightning stroke current characteristic data and aging influence characteristic data to serve as input of a multi-head attention module;
s42, dividing an input sequence into 3 heads, and adding position codes;
S43, performing three different linear transformations on the input sequence added with the position codes to obtain a query Q, a key K and a value V;
S44, normalizing the attention score by using a Softmax function to obtain the attention weight;
s45, carrying out weighted summation on the attention weight and the value V;
s46, fusing the characteristics through the full connection layer, outputting through the linear layer, and outputting a diagnosis result of the lightning arrester.
7. The lightning arrester fault indication judging method according to claim 1, wherein in the training and optimizing process of step S4, the mean square error is used as a loss function, training and optimizing are realized through an optimizer, and an Adam optimizer is selected as an optimizing method of model parameters.
8. A lightning arrester fault indication device is characterized by comprising
A housing mounted on the arrester;
The control center is internally provided with a lightning arrester fault diagnosis model, and is arranged in the shell;
The display screen is connected with the control center and is arranged on the surface of the shell;
an indication fan provided with three blades and having different colors;
The indicating fan is electrically connected with the control center, and rotates to a color corresponding to the diagnosis result when the arrester fault diagnosis model outputs the diagnosis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511438371.2A CN120908582B (en) | 2025-10-10 | 2025-10-10 | Lightning arrester fault indication judging method and indicating device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511438371.2A CN120908582B (en) | 2025-10-10 | 2025-10-10 | Lightning arrester fault indication judging method and indicating device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN120908582A true CN120908582A (en) | 2025-11-07 |
| CN120908582B CN120908582B (en) | 2026-01-30 |
Family
ID=97546423
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202511438371.2A Active CN120908582B (en) | 2025-10-10 | 2025-10-10 | Lightning arrester fault indication judging method and indicating device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120908582B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102818956A (en) * | 2012-08-16 | 2012-12-12 | 大连世有电力科技有限公司 | Lightning arrester online monitoring system with hot-spot detection |
| CN112269142A (en) * | 2020-08-31 | 2021-01-26 | 广东电网有限责任公司 | Lightning arrester detection device and detection method |
| CN112698119A (en) * | 2020-10-14 | 2021-04-23 | 万控智造股份有限公司 | Intelligent instrument device and use method thereof |
| CN115629242A (en) * | 2022-03-01 | 2023-01-20 | 西安西翰电力科技有限公司 | A method and system for predicting, diagnosing and analyzing lightning arrester operating status and damage |
| CN117706179A (en) * | 2023-12-15 | 2024-03-15 | 国网上海市电力公司 | Health state monitoring method of intelligent lightning arrester |
| US20240146047A1 (en) * | 2022-10-31 | 2024-05-02 | Southern States, Llc | Arrester Performance Monitoring and Maintenance System |
| KR102683240B1 (en) * | 2023-04-25 | 2024-07-10 | 가나이엔지 주식회사 | DC lightning arrester system having diagnostic function |
| CN118687621A (en) * | 2024-06-28 | 2024-09-24 | 深圳中鹏新电气技术有限公司 | A method and system for monitoring the real-time operation status of a lightning arrester |
| CN120405338A (en) * | 2025-04-01 | 2025-08-01 | 福建水利电力职业技术学院 | A method for locating the fault point of a lightning arrester |
| CN120405321A (en) * | 2025-06-27 | 2025-08-01 | 成都航空职业技术学院 | A systematic monitoring method for cable electrical safety performance |
-
2025
- 2025-10-10 CN CN202511438371.2A patent/CN120908582B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102818956A (en) * | 2012-08-16 | 2012-12-12 | 大连世有电力科技有限公司 | Lightning arrester online monitoring system with hot-spot detection |
| CN112269142A (en) * | 2020-08-31 | 2021-01-26 | 广东电网有限责任公司 | Lightning arrester detection device and detection method |
| CN112698119A (en) * | 2020-10-14 | 2021-04-23 | 万控智造股份有限公司 | Intelligent instrument device and use method thereof |
| CN115629242A (en) * | 2022-03-01 | 2023-01-20 | 西安西翰电力科技有限公司 | A method and system for predicting, diagnosing and analyzing lightning arrester operating status and damage |
| US20240146047A1 (en) * | 2022-10-31 | 2024-05-02 | Southern States, Llc | Arrester Performance Monitoring and Maintenance System |
| KR102683240B1 (en) * | 2023-04-25 | 2024-07-10 | 가나이엔지 주식회사 | DC lightning arrester system having diagnostic function |
| CN117706179A (en) * | 2023-12-15 | 2024-03-15 | 国网上海市电力公司 | Health state monitoring method of intelligent lightning arrester |
| CN118687621A (en) * | 2024-06-28 | 2024-09-24 | 深圳中鹏新电气技术有限公司 | A method and system for monitoring the real-time operation status of a lightning arrester |
| CN120405338A (en) * | 2025-04-01 | 2025-08-01 | 福建水利电力职业技术学院 | A method for locating the fault point of a lightning arrester |
| CN120405321A (en) * | 2025-06-27 | 2025-08-01 | 成都航空职业技术学院 | A systematic monitoring method for cable electrical safety performance |
Non-Patent Citations (1)
| Title |
|---|
| 张佩;王灵梅;赵兴勇;: "基于BP神经网络氧化锌避雷器的故障诊断研究", 山西电力, no. 04, 15 August 2013 (2013-08-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN120908582B (en) | 2026-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113687255B (en) | A battery cell status diagnosis method, device and storage medium | |
| CN113486291B (en) | Petroleum drilling machine micro-grid fault prediction method based on deep learning | |
| CN114370380A (en) | Wind turbine generator fault diagnosis method considering meteorological factors | |
| CN114977483A (en) | A fault diagnosis system for smart grid regulation and control equipment | |
| CN119535233B (en) | A method for assessing abnormal states of substation battery banks based on multimodal characteristics | |
| CN118798422A (en) | A device and method for predicting power equipment faults based on multimodal perception of power acquisition system | |
| CN117749093A (en) | LSTM-based photovoltaic module quality detection method | |
| CN119355533A (en) | A battery fault diagnosis method based on multi-dimensional feature fusion | |
| CN117686916B (en) | Laboratory digital management system | |
| CN120908582B (en) | Lightning arrester fault indication judging method and indicating device | |
| CN115940809A (en) | Solar panel fault detection method based on power data and visual analysis | |
| CN114997664A (en) | A method and system for early warning analysis of transformer abnormality based on convolutional neural network | |
| CN119813518B (en) | A Deep Learning-Based Intelligent Inspection Management System and Method for Smart Railway Lines | |
| CN121093229A (en) | Power transmission line abnormality detection method and device | |
| CN120820859A (en) | Battery system health status assessment method and system based on multi-dimensional feature fusion | |
| CN119148044B (en) | Ammeter anomaly detection method | |
| CN115184728A (en) | Fault identification method for AC/DC hybrid power distribution network | |
| CN116796261B (en) | Closed switch equipment mechanical characteristic prediction method based on artificial intelligence | |
| Ballal et al. | SPARK: A Practical Framework for Solar PV Panel Diagnostics and Performance Evaluation | |
| CN111915028B (en) | A method for analyzing environmental adaptability of DC charging piles | |
| Josè et al. | Artificial Intelligence for Fault Detection in Photovoltaic Panels | |
| CN120847631B (en) | Battery fault identification method and system | |
| Sun et al. | Fault diagnosis and state evaluation of distributed photovoltaic systems in microgrids | |
| CN121276382B (en) | A method and system for determining battery cell anomalies based on BMS data. | |
| CN120891398B (en) | Lithium battery internal short circuit diagnosis method and system based on double-state circulation equivalent circuit |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |