CN116739996A - Power transmission line insulator fault diagnosis method based on deep learning - Google Patents

Power transmission line insulator fault diagnosis method based on deep learning Download PDF

Info

Publication number
CN116739996A
CN116739996A CN202310585761.7A CN202310585761A CN116739996A CN 116739996 A CN116739996 A CN 116739996A CN 202310585761 A CN202310585761 A CN 202310585761A CN 116739996 A CN116739996 A CN 116739996A
Authority
CN
China
Prior art keywords
insulator
deep learning
fault diagnosis
transmission line
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310585761.7A
Other languages
Chinese (zh)
Inventor
郭保苏
陈昊哲
吴凤和
孙迎兵
张磊
丁世琛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yanshan University
Original Assignee
Yanshan University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Yanshan University filed Critical Yanshan University
Priority to CN202310585761.7A priority Critical patent/CN116739996A/en
Publication of CN116739996A publication Critical patent/CN116739996A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/70—Labelling scene content, e.g. deriving syntactic or semantic representations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
    • Y04S10/52—Outage or fault management, e.g. fault detection or location

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Quality & Reliability (AREA)
  • Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)

Abstract

The application relates to a power transmission line insulator fault diagnosis method based on deep learning, which comprises the following steps of: acquiring and processing image information and working condition information of an insulator of a power transmission line; step 2: training an insulator fault diagnosis model according to the insulator image data and the insulator working condition characteristic matrix data; step 3: determining the training effect of a fault diagnosis model of the insulator of the power transmission line, and storing the trained model; step 4: and applying the insulator fault diagnosis model to online fault diagnosis of the insulator of the power transmission line. The deep learning-based power transmission line insulator fault diagnosis method provided by the application can accurately classify different fault types, greatly improve the power transmission line insulator fault diagnosis speed, realize end-to-end fault diagnosis of the power transmission line insulator through integrating a data preprocessing process and an end-to-end deep learning network, simplify a fault diagnosis process and obtain higher fault diagnosis accuracy.

Description

Power transmission line insulator fault diagnosis method based on deep learning
Technical Field
The application relates to the technical field related to fault diagnosis of physical elements, in particular to a power transmission line insulator fault diagnosis method based on deep learning.
Background
Insulators (insulators) are physical elements that are mounted between conductors or conductors of different electrical potential and a grounded member, and are capable of withstanding voltage and mechanical stress. Insulators are various in variety and shape. Different types of insulators have large differences in structure and appearance, but are composed of two major parts, namely an insulating part and a connecting fitting. The insulator of the power transmission line is an original part with the largest share in the whole power system, meanwhile, the fault of the power transmission line is directly related to the power transmission line to a great extent, and the insulator fault directly or indirectly causes most of faults of the power transmission line. The types of insulators of the power transmission line are many, and according to the different materials of the insulators, the insulators can be divided into three main types of porcelain insulators, glass insulators, composite insulators and the like, and the types of faults which are easy to occur in the insulators of different types are obviously different. In the long-term operation of the power transmission line, the insulator is affected by a lot of external environments, such as climate influence, temperature change and the like, and the state of the insulator is affected by the difference of current and voltage of the power transmission line; in addition, the mechanical loading of the insulator can also affect the state of the insulator. At present, the possible fault types of the power transmission line insulator under various working conditions can be summarized into six types of arc climbing, string falling, self-explosion, fracture, resistance degradation and surface pollution.
The traditional transmission line insulator fault diagnosis method comprises the following steps: the spark gap method is based on the fact that whether the insulator can generate discharge or not is a big basis for judging whether the insulator has faults or not; the small ball discharging method is used for carrying out fault diagnosis by measuring small balls at two ends of an insulator and observing the discharging distance of the small balls aiming at the voltage distribution of the insulator, and the two methods have the defects that the specific fault type cannot be judged and the diagnosis accuracy is low; the thermal infrared imager method mainly uses the principle of thermal effect on the surface of an insulator, the surface temperature of a fault insulator is lower than that of a normal insulator, and the method has the defect that the specific fault type cannot be diagnosed although the accuracy is higher; according to the leakage current detection method, leakage current flowing through two ends of the insulator can be measured through the current sensor, so that fault diagnosis of the insulator is achieved, but live detection cannot be achieved through the method, and diagnosis cost is too high. The fault diagnosis technology based on deep learning visual identification is rapid in development and is also applied to the field of power transmission line insulator fault diagnosis, but the existing method only realizes diagnosis of various fault categories through the insulator images, does not consider the influence of working conditions on insulator faults, and cannot detect faults with insignificant visual changes such as resistance degradation.
As the application of the attention mechanism network in the deep learning becomes wider, a visual question-answering task network combining an attention mechanism and a deep convolutional neural network is developed; compared with a diagnosis method which simply depends on images, the deep neural network based on the attention mechanism and the convolutional neural network can realize the combined analysis of the image characteristics and the insulator working condition characteristics, and has more advantages in the field of insulator fault diagnosis.
Disclosure of Invention
In order to overcome the defects of the prior art, the deep learning-based power transmission line insulator fault diagnosis method provided by the application can accurately classify different fault types, greatly improve the power transmission line insulator fault diagnosis speed, realize the end-to-end fault diagnosis of the power transmission line insulator by integrating a data preprocessing process and an end-to-end deep learning network, simplify the fault diagnosis process and obtain higher fault diagnosis accuracy.
In order to achieve the above object, the solution adopted by the present application is:
a power transmission line insulator fault diagnosis method based on deep learning comprises the following steps:
step 1: acquiring and processing image information and working condition information of an insulator of a power transmission line;
acquiring image information of an insulator of the power transmission line, performing image scale adjustment through pixel sampling, and performing normalization processing to obtain data of the insulator image;
collecting working condition information of an insulator of a power transmission line, wherein the working condition information comprises 8-dimensional data of insulator materials, power transmission voltage, power transmission current, mechanical load of the insulator, steel cap temperature, insulator temperature, environmental temperature and weather conditions, the 8-dimensional data are standardized, and uniform-dimensional insulator working condition characteristic matrix data are obtained through background data filling;
step 2: training an insulator fault diagnosis model according to the insulator image data and the insulator working condition characteristic matrix data;
constructing training data according to the insulator image data and the insulator working condition characteristic matrix data in the step 1, dividing the training data into a training data set and a verification data set according to a proportion, and transmitting the training data set into a power transmission line insulator fault diagnosis model based on deep learning for training; the deep learning-based power transmission line insulator fault diagnosis model training comprises the following steps: the system comprises an ICN deep learning module, an ECNN deep learning module, a TSAN deep learning module, a common attention mechanism layer and an output full-connection layer;
the key model structure of the TSAN deep learning module is a self-attention mechanism layer, and the self-attention mechanism layer expression is as follows:
wherein: attention (Q, K, V) represents a self-Attention mechanism function; q represents first intermediate data of the self-attention mechanism layer; k represents second intermediate data of the self-attention mechanism layer; v represents third intermediate data of the self-attention mechanism layer; swish represents the first activation function of the self-attention mechanism layer; w (W) i A first learning parameter representing a self-attention mechanism layer; b i A second learning parameter representing a self-attention mechanism layer; d represents the vector length of the first intermediate data Q and the second intermediate data K of the self-attention mechanism layer; x represents the self-attention mechanism layer input of the TSAN deep learning module; i represents different parameter numbers;
the common attention mechanism layer can realize the fusion of the characteristics of the insulator sub-images and the semantic characteristics of the working conditions of the insulators, and the common attention module trains together with the characteristic extraction network and automatically optimizes learning parameters; the expression of the common attention mechanism layer is as follows:
wherein: alpha represents first intermediate data of the common attention mechanism layer; u represents a first learning parameter of the common attention mechanism layer; y represents the deep learning module output of the ICN network and the ECNN network; l represents a second learning parameter of the common attention mechanism layer; beta represents second intermediate data of the common attention mechanism layer; sigmoid represents a common attention mechanism layer first activation function; swishb represents a second activation function of the common attention mechanism layer; z represents the self-attention mechanism layer output of the TSAN deep learning module; m represents a fourth learning parameter of the common attention mechanism layer; output represents the common attention mechanism layer output;
the activation function in the output full-connection layer is softmax, and the specific expression is as follows:
wherein: j represents the full connection layer neuron number; c (C) j An output representing a j-th neuron; omega j A first learning parameter representing an output full connection layer; beta j A second learning parameter representing an output full connection layer; class of things j Representing a probability that the input data belongs to a j-th defect class; * Representing a matrix multiplication; softmax represents the activation function in the output fully connected layer;
step 3: determining the training effect of a fault diagnosis model of the insulator of the power transmission line, and storing the trained model;
judging the training effect of the transmission line insulator fault diagnosis model according to the verification data set in the step 2, outputting the fault type of the abnormal insulator based on the transmission line insulator fault diagnosis model subjected to deep learning, finishing training by the model when the average absolute error of the verification data set is less than 0.9%, and storing the trained transmission line insulator fault diagnosis model parameters, wherein the average absolute error calculation formula is as follows:
wherein: lmp represents the average absolute error of the dataset; n represents the number of data sets batch; k represents a batch number; ACC (ACC) k Representing the absolute accuracy of the network reasoning result in the kth batch;
step 4: applying an insulator fault diagnosis model to online fault diagnosis of the insulator of the power transmission line;
the input data of the on-line fault diagnosis firstly needs to be subjected to the data preprocessing operation which is the same as that of the training data in the step 1, and the data is transmitted into an insulator fault diagnosis model based on deep learning to obtain the fault type of the insulator of the power transmission line, and finally the fault diagnosis of the insulator of the power transmission line is completed.
Preferably, the training data in the step 2 needs to use the fault diagnosis result of the professional transmission line engineer to make a real fault condition label, and the fault condition of the insulator includes 6 categories of arc climbing, string falling, self-explosion, fracture, resistance degradation and surface pollution.
Preferably, the feature extraction network of the deep learning-based transmission line insulator fault diagnosis model in the step 2 includes an ICN deep learning module and an ECNN deep learning module and a TSAN deep learning module, and a common attention mechanism layer and an output full connection layer, each of which is composed of a convolution layer, a deconvolution layer, a Concat mechanism, an activation function, a pooling layer and a full connection layer.
Preferably, the deep learning-based transmission line insulator fault diagnosis model in step 2 needs to calculate a cross entropy loss function, as follows:
wherein: l represents a cross entropy loss function; m represents the total number of fault categories; y is i Representing the confidence of the ith fault class; y is i Indicating whether it is actually the fault category.
Preferably, the insulator image data in the step 2 needs to be input into an ICN network for feature integration, the output data of the ICN network is input into an ECNN network to obtain insulator image feature data, the insulator working condition feature data is input into a TSAN deep learning module to obtain insulator working condition semantic feature data, the insulator image feature data and the insulator working condition semantic feature data are transmitted into a common attention mechanism module, and finally a diagnosis result is output through an output full-connection layer;
preferably, the ICN deep learning module and the ECNN deep learning module in the step 2 specifically are:
the ICN deep learning module uses three downsampling convolution layers and a maximum value pooling with a step length of 2, and deconvolution layers corresponding to downsampling convolution;
the ECNN deep learning module comprises two convolution layers, wherein the convolution kernel sizes of the two convolution layers are 5, the convolution step sizes are 1, a maximum value pooling series structure with the two step sizes of 2 is used, and the total number of the two pooling layers is four.
Preferably, in the common attention mechanism layer in step 2, the influence degree of the insulator working condition information is obtained through the full connection layer and the Sigmoid function, and then the real image feature data of the insulator obtained by the ICN and ECNN deep learning module is combined to obtain the affected integrated feature data, and the fault diagnosis result is output through the Softmax function.
Preferably, the TSAN deep learning module uses a multi-head self-attention mechanism with the head number of 8, a feature unfolding layer and a full connection layer; the TSAN deep learning module can automatically extract fault diagnosis sensitive information of insulator working condition data, and the TSAN training together with the whole network can automatically optimize learning parameters; the diagnosis sensitive information of the insulator working condition data obtained by the TSAN deep learning module can influence a final fault diagnosis result through the common attention mechanism module.
Compared with the prior art, the application has the beneficial effects that:
(1) The network model provided by the application uses the insulator working condition data characteristic extraction network based on the self-attention mechanism, and can perform parameter optimization along with the whole network, so as to realize quantification of influence of the insulator working condition on the diagnosis result; the fault diagnosis result output network based on the common attention mechanism is used, and the network output module can perform parameter optimization together with the whole network, so that the correlation between the insulator working condition characteristic influence and the insulator sub-image characteristic and the result output are realized.
(2) The power transmission line insulator fault diagnosis method based on deep learning provided by the application can greatly improve the power transmission line insulator fault diagnosis speed, accurately classify different fault types, provide references for subsequent fault resolution and elimination, and help to quickly restore power supply. The deep learning-based power transmission line insulator fault diagnosis method provided by the application realizes the end-to-end fault diagnosis of the power transmission line insulator through integrating the data preprocessing process and the end-to-end deep learning network.
(3) The application provides a more convenient diagnosis mode for a user, can operate without mastering a great deal of professional power transmission engineering knowledge, simplifies the insulator fault diagnosis process, and enables a common worker to realize insulator fault diagnosis under the condition of completing data acquisition; by using the designed network structure, higher fault diagnosis accuracy is realized than that of the existing deep learning network.
Drawings
Fig. 1 is a control block diagram of a deep learning-based transmission line insulator fault diagnosis method according to an embodiment of the present application;
fig. 2 is a flow chart of steps of a power transmission line insulator fault diagnosis method based on deep learning according to an embodiment of the application;
FIG. 3 is a diagram of a network model architecture according to an embodiment of the present application;
FIG. 4 is a computational graph of a common attention mechanism layer in a network model according to an embodiment of the present application;
FIG. 5 is a computational graph of a multi-headed self-attention layer in a network model according to an embodiment of the present application;
FIG. 6 is a graph showing the change of the loss value and the accuracy rate in the network training process according to the embodiment of the application;
FIG. 7 is a diagram of a confusion matrix for network reasoning results that considers only two fault types in accordance with an embodiment of the present application;
fig. 8 is a diagram showing an example of a real image of an insulator used in an embodiment of the present application.
Detailed Description
Hereinafter, embodiments of the present application will be described with reference to the drawings.
The network model provided by the embodiment of the application uses an insulator working condition data characteristic extraction network based on a self-attention mechanism to realize quantification of influence of insulator working conditions on diagnosis results, and uses a fault diagnosis result output network based on a common attention mechanism to realize association and result output of the influence of insulator working condition characteristics and insulator sub-image characteristics; the power transmission line insulator fault diagnosis method based on deep learning can greatly improve the power transmission line insulator fault diagnosis speed, accurately classify different fault types, provide references for subsequent fault resolution and elimination, and help to quickly restore power supply. The deep learning-based power transmission line insulator fault diagnosis method provided by the application realizes the end-to-end fault diagnosis of the power transmission line insulator through integrating the data preprocessing process and the end-to-end deep learning network; the case simplifies the insulator fault diagnosis process, enables a common worker to realize insulator fault diagnosis under the condition of completing data acquisition, and improves the fault diagnosis accuracy. Fig. 1 is a control block diagram of a deep learning-based transmission line insulator fault diagnosis method according to an embodiment of the present application.
The embodiment of the application provides a power transmission line insulator fault diagnosis method based on deep learning, and as shown in fig. 2, the power transmission line insulator fault diagnosis method based on deep learning in the embodiment of the application comprises the steps of a flow chart; to demonstrate the applicability of the application, it is applied to examples, comprising in particular the following steps:
s1: acquiring and processing image information and working condition information of an insulator of a power transmission line;
image information of an insulator of the power transmission line is collected, image scale adjustment is carried out through pixel sampling, normalization processing is carried out, insulator image data are obtained, and an input image of a network is shown in fig. 8 and is an example diagram of a real image of the insulator used in the embodiment of the application.
Working condition information of an insulator of a power transmission line is collected, wherein the working condition information comprises 8-dimensional data including insulator materials, power transmission voltage, power transmission current, mechanical load of the insulator, steel cap temperature, insulator temperature, environmental temperature and weather conditions, the 8-dimensional data are standardized, uniform-dimension insulator working condition characteristic matrix data are obtained through background data filling, and the insulator working condition data are one-dimensional arrays containing eight data elements.
S2: training an insulator fault diagnosis model according to the insulator image data and the insulator working condition characteristic matrix data;
constructing training data according to the insulator image data and the insulator working condition characteristic matrix data in the step S1, dividing the training data into a training data set and a verification data set according to a proportion, and transmitting the training data set into a power transmission line insulator fault diagnosis model based on deep learning for training; the training data is needed to manufacture a real fault condition label by using a fault diagnosis result of a professional transmission line engineer, the fault condition of the insulator comprises 6 categories of arc climbing, string falling, self explosion, fracture, resistance degradation and surface pollution, and the category label of the data set is given by a six-dimensional single-heat coding mode.
The training of the transmission line insulator fault diagnosis model based on deep learning comprises the following steps: the system comprises an ICN deep learning module, an ECNN deep learning module, a TSAN deep learning module, a common attention mechanism layer and an output full-connection layer; FIG. 4 is a computational graph of the common attention mechanism layer in the network model according to an embodiment of the present application. The feature extraction network of the transmission line insulator fault diagnosis model based on deep learning comprises an ICN deep learning module, an ECNN deep learning module, a TSAN deep learning module, a common attention mechanism layer and an output full-connection layer, wherein the ICN deep learning module consists of a convolution layer, a deconvolution layer, a Concat mechanism, an activation function, a pooling layer and a full-connection layer. FIG. 3 is a diagram of a network model structure according to an embodiment of the present application. The TSAN deep learning module uses a multi-head self-attention mechanism with the head number of 8, a characteristic unfolding layer and a full connection layer; FIG. 5 is a diagram illustrating the calculation of the multi-headed self-attention layer in the network model according to the embodiment of the present application. The TSAN deep learning module can automatically extract fault diagnosis sensitive information of insulator working condition data, and the TSAN training together with the whole network can automatically optimize learning parameters; the diagnosis sensitive information of the insulator working condition data obtained by the TSAN deep learning module can influence the final fault diagnosis result through the common attention mechanism module, and when the fault type is 6 and the batch size is set to 2, one example of the final result output by the network is as follows:
and then, converting the output data into fault category output by using the conditional statement, and completing the model reasoning process.
The method comprises the steps that insulator image data are required to be input into an ICN (information and communication network) for feature integration, output data of the ICN are input into an ECNN to obtain insulator image feature data, insulator working condition feature data are input into a TSAN deep learning module to obtain insulator working condition semantic feature data, then the insulator image feature data and the insulator working condition semantic feature data are transmitted into a common attention mechanism module, and finally a diagnosis result is output through an output full-connection layer; the ICN deep learning module uses three downsampling convolution layers and a maximum value pooling with a step length of 2 and a deconvolution layer corresponding to the downsampling convolution; the ECNN deep learning module comprises two convolution layers, wherein the convolution kernel sizes of the two convolution layers are 5, the convolution step sizes are 1, a maximum value pooling serial structure with the two step sizes of 2 is used, and the total number of the two pooling layers is four.
The common attention mechanism layer obtains the influence degree of the insulator working condition information through the full connection layer and the Sigmoid function, and then combines the insulator real image characteristic data obtained by the ICN and ECNN deep learning module to obtain the affected integrated characteristic data and outputs a fault diagnosis result through the Softmax function.
The key model structure of the TSAN deep learning module is a self-attention mechanism layer, and the expression of the self-attention mechanism layer is as follows:
wherein: attention (Q, K, V) represents a self-Attention mechanism function; q represents first intermediate data of the self-attention mechanism layer; k represents second intermediate data of the self-attention mechanism layer; v represents third intermediate data of the self-attention mechanism layer; swish represents the first activation function of the self-attention mechanism layer; w (W) i A first learning parameter representing a self-attention mechanism layer; b i A second learning parameter representing a self-attention mechanism layer; d represents the vector length of the first intermediate data Q and the second intermediate data K of the self-attention mechanism layer; x represents the self-attention mechanism layer input of the TSAN deep learning module; i represents different parameter numbers.
The common attention mechanism layer can realize the fusion of the characteristics of the insulator sub-images and the semantic characteristics of the working conditions of the insulators, and the common attention module trains together with the characteristic extraction network and automatically optimizes learning parameters; the expression of the common attention mechanism layer is as follows:
wherein: alpha represents first intermediate data of the common attention mechanism layer; u represents a first learning parameter of the common attention mechanism layer; y represents the deep learning module output of the ICN network and the ECNN network; l represents a second learning parameter of the common attention mechanism layer; beta represents second intermediate data of the common attention mechanism layer; sigmoid represents a common attention mechanism layer first activation function; swishb represents a second activation function of the common attention mechanism layer; z represents the self-attention mechanism layer output of the TSAN deep learning module; m represents a fourth learning parameter of the common attention mechanism layer; output represents the common attention mechanism layer output.
The activation function in the output full connection layer is softmax, and the specific expression is as follows:
wherein: j represents the full connection layer neuron number; c (C) j An output representing a j-th neuron; omega j A first learning parameter representing an output full connection layer; beta j A second learning parameter representing an output full connection layer; class of things j Representing a probability that the input data belongs to a j-th defect class; * Representing a matrix multiplication; softmax represents the activation function in the output fully connected layer.
The transmission line insulator fault diagnosis model based on deep learning needs to calculate a cross entropy loss function, and as shown in fig. 6, a loss value and accuracy rate change chart in the network training process in the embodiment of the application; the following is shown:
wherein: l represents a cross entropy loss function; m represents the total number of fault categories; y is i Indicating the ith faultCategory confidence; y is i Indicating whether it is actually the fault category.
S3: determining the training effect of a fault diagnosis model of the insulator of the power transmission line, and storing the trained model;
judging the training effect of the transmission line insulator fault diagnosis model according to the verification data set in the S2, outputting the fault type of the abnormal insulator based on the transmission line insulator fault diagnosis model subjected to deep learning, finishing training by the model when the average absolute error of the verification data set is less than 0.9%, and storing the trained transmission line insulator fault diagnosis model parameters, wherein the average absolute error calculation formula is as follows:
wherein: lmp represents the average absolute error of the dataset; n represents the number of data sets batch; k represents a batch number; ACC (ACC) k Representing the absolute accuracy of the network reasoning results in the kth batch.
S4: applying an insulator fault diagnosis model to online fault diagnosis of the insulator of the power transmission line;
the input data of the on-line fault diagnosis firstly needs to be subjected to the data preprocessing operation which is the same as that of the training data in the step S1, and the data is transmitted into an insulator fault diagnosis model based on deep learning to obtain the fault type of the insulator of the power transmission line, and finally the fault diagnosis of the insulator of the power transmission line is completed. Fig. 7 shows a confusion matrix diagram of network reasoning results only considering two fault types, and analysis of the confusion matrix can show that the application has a good effect on fault diagnosis of the insulator of the power transmission line, and can meet actual use requirements.
In conclusion, the prediction result of the transmission line insulator fault diagnosis method based on deep learning proves that the method has a good effect.
(1) According to the embodiment of the application, the actual power transmission line data is processed through the insulator working condition data characteristic extraction network based on the self-attention mechanism, so that parameter optimization can be carried out along with the whole network, and the quantification of the influence of the insulator working condition on the diagnosis result is realized; the fault diagnosis result output network based on the common attention mechanism is used, and the network output module can perform parameter optimization together with the whole network, so that the correlation between the insulator working condition characteristic influence and the insulator sub-image characteristic and the result output are realized.
(2) The power transmission line insulator fault diagnosis method based on deep learning provided by the embodiment of the application can greatly improve the power transmission line insulator fault diagnosis speed, accurately classify different fault types, provide references for subsequent fault resolution and elimination, and help to quickly restore power supply. The deep learning-based power transmission line insulator fault diagnosis method provided by the application realizes the end-to-end fault diagnosis of the power transmission line insulator through integrating the data preprocessing process and the end-to-end deep learning network.
(3) The embodiment of the application simplifies the insulator fault diagnosis process, can operate without mastering a large amount of professional power transmission engineering knowledge, enables a common worker to realize insulator fault diagnosis under the condition of completing data acquisition, and can be seen through comparison of diagnosis results.
The above examples are only illustrative of the preferred embodiments of the present application and are not intended to limit the scope of the present application, and various modifications and improvements made by those skilled in the art to the technical solution of the present application should fall within the scope of protection defined by the claims of the present application without departing from the spirit of the present application.

Claims (8)

1. The power transmission line insulator fault diagnosis method based on deep learning is characterized by comprising the following steps of:
step 1: acquiring and processing image information and working condition information of an insulator of a power transmission line;
acquiring image information of an insulator of the power transmission line, performing image scale adjustment through pixel sampling, and performing normalization processing to obtain data of the insulator image;
collecting working condition information of an insulator of a power transmission line, wherein the working condition information comprises 8-dimensional data of insulator materials, power transmission voltage, power transmission current, mechanical load of the insulator, steel cap temperature, insulator temperature, environmental temperature and weather conditions, the 8-dimensional data are standardized, and uniform-dimensional insulator working condition characteristic matrix data are obtained through background data filling;
step 2: training an insulator fault diagnosis model according to the insulator image data and the insulator working condition characteristic matrix data;
constructing training data according to the insulator image data and the insulator working condition characteristic matrix data in the step 1, dividing the training data into a training data set and a verification data set according to a proportion, and transmitting the training data set into a power transmission line insulator fault diagnosis model based on deep learning for training; the deep learning-based power transmission line insulator fault diagnosis model training comprises the following steps: the system comprises an ICN deep learning module, an ECNN deep learning module, a TSAN deep learning module, a common attention mechanism layer and an output full-connection layer;
the key model structure of the TSAN deep learning module is a self-attention mechanism layer, and the self-attention mechanism layer expression is as follows:
wherein: attention (Q, K, V) represents a self-Attention mechanism function; q represents first intermediate data of the self-attention mechanism layer; k represents second intermediate data of the self-attention mechanism layer; v represents third intermediate data of the self-attention mechanism layer; swish represents the first activation function of the self-attention mechanism layer; w (W) i A first learning parameter representing a self-attention mechanism layer; b i A second learning parameter representing a self-attention mechanism layer; d represents the vector length of the first intermediate data Q and the second intermediate data K of the self-attention mechanism layer; x represents the self-attention mechanism layer input of the TSAN deep learning module; i represents different parameter numbers;
the common attention mechanism layer can realize the fusion of the characteristics of the insulator sub-images and the semantic characteristics of the working conditions of the insulators, and the common attention module trains together with the characteristic extraction network and automatically optimizes learning parameters; the expression of the common attention mechanism layer is as follows:
wherein: alpha represents first intermediate data of the common attention mechanism layer; u represents a first learning parameter of the common attention mechanism layer; y represents the deep learning module output of the ICN network and the ECNN network; l represents a second learning parameter of the common attention mechanism layer; beta represents second intermediate data of the common attention mechanism layer; sigmoid represents a common attention mechanism layer first activation function; swishb represents a second activation function of the common attention mechanism layer; z represents the self-attention mechanism layer output of the TSAN deep learning module; m represents a fourth learning parameter of the common attention mechanism layer; output represents the common attention mechanism layer output;
the activation function in the output full-connection layer is softmax, and the specific expression is as follows:
wherein: j represents the full connection layer neuron number; c (C) j An output representing a j-th neuron; omega j A first learning parameter representing an output full connection layer; beta j A second learning parameter representing an output full connection layer; class of things j Representing a probability that the input data belongs to a j-th defect class; * Representing a matrix multiplication; softmax represents the activation function in the output fully connected layer;
step 3: determining the training effect of a fault diagnosis model of the insulator of the power transmission line, and storing the trained model;
judging the training effect of the transmission line insulator fault diagnosis model according to the verification data set in the step 2, outputting the fault type of the abnormal insulator based on the transmission line insulator fault diagnosis model subjected to deep learning, finishing training by the model when the average absolute error of the verification data set is less than 0.9%, and storing the trained transmission line insulator fault diagnosis model parameters, wherein the average absolute error calculation formula is as follows:
wherein: lmp represents the average absolute error of the dataset; n represents the number of data sets batch; k represents a batch number; ACC (ACC) k Representing the absolute accuracy of the network reasoning result in the kth batch;
step 4: applying an insulator fault diagnosis model to online fault diagnosis of the insulator of the power transmission line;
the input data of the on-line fault diagnosis firstly needs to be subjected to the data preprocessing operation which is the same as that of the training data in the step 1, and the data is transmitted into an insulator fault diagnosis model based on deep learning to obtain the fault type of the insulator of the power transmission line, and finally the fault diagnosis of the insulator of the power transmission line is completed.
2. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the training data in the step 2 is required to make a real fault condition label by using a fault diagnosis result of a professional power transmission line engineer, and the fault condition of the insulator includes 6 categories of arc climbing, string falling, self-explosion, fracture, resistance degradation and surface pollution.
3. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the feature extraction network of the deep learning-based power transmission line insulator fault diagnosis model in step 2 comprises an ICN deep learning module and an ECNN deep learning module and a TSAN deep learning module, a common attention mechanism layer and an output full connection layer, each of which is composed of a convolution layer, a deconvolution layer, a Concat mechanism, an activation function, a pooling layer and a full connection layer.
4. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the deep learning-based power transmission line insulator fault diagnosis model in step 2 needs to calculate a cross entropy loss function as follows:
wherein: l represents a cross entropy loss function; m represents the total number of fault categories; y is i Representing the confidence of the ith fault class; y is i Indicating whether it is actually the fault category.
5. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the insulator image data in the step 2 is required to be input into an ICN network for feature integration, the output data of the ICN network is input into the ECNN network to obtain insulator image feature data, the insulator condition feature data is input into a TSAN deep learning module to obtain insulator condition semantic feature data, the insulator image feature data and the insulator condition semantic feature data are input into a common attention mechanism module, and finally a diagnosis result is output through an output full connection layer.
6. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the ICN deep learning module and the ECNN deep learning module in step 2 specifically are:
the ICN deep learning module uses three downsampling convolution layers and a maximum value pooling with a step length of 2, and deconvolution layers corresponding to downsampling convolution;
the ECNN deep learning module comprises two convolution layers, wherein the convolution kernel sizes of the two convolution layers are 5, the convolution step sizes are 1, a maximum value pooling series structure with the two step sizes of 2 is used, and the total number of the two pooling layers is four.
7. The deep learning-based power transmission line insulator fault diagnosis method according to claim 1, wherein the degree of influence of the insulator working condition information is obtained through a full connection layer and a Sigmoid function in the common attention mechanism layer in the step 2, and then the actual image feature data of the insulator obtained by the ICN and ECNN deep learning module is combined to obtain the affected integrated feature data and output a fault diagnosis result through a Softmax function.
8. The deep learning-based transmission line insulator fault diagnosis method according to claim 3, wherein a multi-head self-attention mechanism with the number of heads of 8, a characteristic unfolding layer and a full connection layer are used in the TSAN deep learning module; the TSAN deep learning module can automatically extract fault diagnosis sensitive information of insulator working condition data, and the TSAN training together with the whole network can automatically optimize learning parameters; the diagnosis sensitive information of the insulator working condition data obtained by the TSAN deep learning module can influence a final fault diagnosis result through the common attention mechanism module.
CN202310585761.7A 2023-05-23 2023-05-23 Power transmission line insulator fault diagnosis method based on deep learning Pending CN116739996A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310585761.7A CN116739996A (en) 2023-05-23 2023-05-23 Power transmission line insulator fault diagnosis method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310585761.7A CN116739996A (en) 2023-05-23 2023-05-23 Power transmission line insulator fault diagnosis method based on deep learning

Publications (1)

Publication Number Publication Date
CN116739996A true CN116739996A (en) 2023-09-12

Family

ID=87907088

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310585761.7A Pending CN116739996A (en) 2023-05-23 2023-05-23 Power transmission line insulator fault diagnosis method based on deep learning

Country Status (1)

Country Link
CN (1) CN116739996A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117011805A (en) * 2023-10-07 2023-11-07 广东电网有限责任公司云浮供电局 Data exception evaluation method, device, equipment and readable storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117011805A (en) * 2023-10-07 2023-11-07 广东电网有限责任公司云浮供电局 Data exception evaluation method, device, equipment and readable storage medium
CN117011805B (en) * 2023-10-07 2024-02-06 广东电网有限责任公司云浮供电局 Data exception evaluation method, device, equipment and readable storage medium

Similar Documents

Publication Publication Date Title
CN110598736B (en) Power equipment infrared image fault positioning, identifying and predicting method
CN110533631B (en) SAR image change detection method based on pyramid pooling twin network
CN110376522B (en) Motor fault diagnosis method of data fusion deep learning network
CN110895814B (en) Aero-engine hole-finding image damage segmentation method based on context coding network
CN111382542A (en) Road electromechanical equipment life prediction system oriented to full life cycle
CN111161224A (en) Casting internal defect grading evaluation system and method based on deep learning
CN116739996A (en) Power transmission line insulator fault diagnosis method based on deep learning
CN117809164A (en) Substation equipment fault detection method and system based on multi-mode fusion
Zheng et al. Intelligent diagnosis method of power equipment faults based on single‐stage infrared image target detection
CN116612098A (en) Insulator RTV spraying quality evaluation method and device based on image processing
CN115147591A (en) Transformer equipment infrared image voltage heating type defect diagnosis method and system
CN112418324B (en) Cross-modal data fusion method for electrical equipment state perception
CN114487643A (en) On-spot handing-over of extra-high voltage GIL equipment is accepted and is synthesized test platform
CN116485802B (en) Insulator flashover defect detection method, device, equipment and storage medium
CN117173595A (en) Unmanned aerial vehicle aerial image target detection method based on improved YOLOv7
CN116738355A (en) Electric power multi-mode feature level fusion method considering data difference
CN113554070B (en) Bearing fault diagnosis method based on transition probability matrix and lightweight network
CN115471453A (en) Power equipment fault diagnosis method based on image processing
CN113344915A (en) Method and system for detecting defects of key parts of power transmission line
CN115017984A (en) Early warning method and system for failure risk of aircraft engine
CN114387524A (en) Image identification method and system for small sample learning based on multilevel second-order representation
Bao et al. Automatic identification and defect diagnosis of transmission line insulators based on YOLOv3 network
CN112634216B (en) Insulator self-explosion detection method based on deep learning model
Zaripova Convolutional Neural Network to Detect Support and Suspend Insulators on Infrared Images
CN114972812B (en) Non-local attention mechanics learning method based on structural similarity

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