CN113971777A - Equipment fault prediction method and device and server - Google Patents

Equipment fault prediction method and device and server Download PDF

Info

Publication number
CN113971777A
CN113971777A CN202111240270.6A CN202111240270A CN113971777A CN 113971777 A CN113971777 A CN 113971777A CN 202111240270 A CN202111240270 A CN 202111240270A CN 113971777 A CN113971777 A CN 113971777A
Authority
CN
China
Prior art keywords
target
prediction
sample
data set
picture
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
CN202111240270.6A
Other languages
Chinese (zh)
Inventor
梅峰
刘绪颖
沈桂竹
王雪
周果清
王庆
姚一杨
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.)
Northwestern Polytechnical University
State Grid Zhejiang Electric Power Co Ltd
Information and Telecommunication Branch of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
Northwestern Polytechnical University
State Grid Zhejiang Electric Power Co Ltd
Information and Telecommunication Branch of State Grid Zhejiang Electric Power Co Ltd
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 Northwestern Polytechnical University, State Grid Zhejiang Electric Power Co Ltd, Information and Telecommunication Branch of State Grid Zhejiang Electric Power Co Ltd filed Critical Northwestern Polytechnical University
Priority to CN202111240270.6A priority Critical patent/CN113971777A/en
Publication of CN113971777A publication Critical patent/CN113971777A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR 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
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Human Resources & Organizations (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Software Systems (AREA)
  • Marketing (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Game Theory and Decision Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Development Economics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a device fault prediction method, a device and a server, which are applied to the technical field of power systems. The method can predict the fault of the target equipment based on the equipment pictures of different target spectra, further determine whether the power equipment is possible to have the fault in advance, provide reference basis for operation and maintenance personnel to take targeted operation and maintenance measures, and solve the problems in the prior art.

Description

Equipment fault prediction method and device and server
Technical Field
The invention belongs to the technical field of power systems, and particularly relates to a method and a device for predicting equipment faults and a server.
Background
With the increase of national power demand, the construction scale of power grids is continuously enlarged, and the power load is increased day by day. Power equipment in a power grid is in a working state of high voltage, high temperature and high load for a long time, and once faults occur in the operation, huge economic loss can be caused, and great potential safety hazards also exist.
In the existing application, the overhaul of the power equipment mostly comprises two conditions, one is the overhaul after the power equipment breaks down, and the other is the overhaul performed according to the preset overhaul period. The maintenance is carried out after the power equipment fails, so that the economic loss caused by power failure is inevitable, and the power equipment in a failure state may bring certain safety risk to maintenance personnel; on the other hand, the maintenance is performed according to the predetermined maintenance period, which may cause waste of manpower and material resources, and on the other hand, the actual application requirements are difficult to meet due to the sudden failure of the equipment.
Based on the above situation, how to predict the failure of the power equipment, and determine in advance whether the power equipment is likely to fail, provide a reference basis for taking targeted operation and maintenance measures, and become one of the technical problems to be solved by those skilled in the art.
Disclosure of Invention
In view of this, an object of the present invention is to provide a device fault prediction method, apparatus and server, which perform fault prediction on a target device based on device pictures of different target spectra, determine in advance whether a power device may have a fault, and provide a reference basis for operation and maintenance personnel to take targeted operation and maintenance measures. The specific scheme is as follows:
in a first aspect, the present invention provides an apparatus failure prediction method, including:
acquiring a device picture set of a target device at a plurality of sampling moments within a preset time interval;
wherein the device picture set comprises device pictures of a plurality of target spectral bands;
extracting reference values of target parameters of target spectrum device pictures in the device picture sets to obtain reference data sets corresponding to the sampling moments;
inputting each reference data set into a pre-trained data prediction model to obtain prediction data sets of the target equipment at a plurality of prediction moments;
wherein the prediction data set comprises a prediction value of each target parameter;
and inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
Optionally, the obtaining of the device picture sets of the target device at a plurality of sampling moments within a preset time interval includes:
acquiring a preset time interval;
determining a plurality of sampling moments in a preset time interval according to the preset time interval;
and respectively acquiring a device picture set of the target device at each sampling moment.
Optionally, the extracting a reference value of a target parameter of each target spectrum device picture in each device picture set includes:
extracting actual parameter values of target parameters of target spectrum device pictures in the device picture sets to obtain actual data sets corresponding to the device picture sets;
processing each actual data set according to a preset preprocessing rule to obtain a corresponding preprocessed actual data set;
the preprocessed actual data set comprises preprocessed actual parameter values of the target parameters;
and respectively carrying out standardization processing on each preprocessed actual parameter value in each preprocessed actual data set to obtain a reference value of a target parameter of each target spectrum device picture in each device picture set.
Optionally, the process of training the data prediction model includes:
acquiring a sample picture set of a plurality of sampling moments of sample equipment in a target time interval;
wherein the sample picture set comprises sample pictures of a plurality of target spectral bands;
extracting sample values of target parameters of each target spectrum band sample picture in each sample picture set to obtain a sample data set of each sampling moment;
dividing each sample data set into a training sample data set and a verification sample data set;
respectively determining the output result of the LSTM neural network to each training sample data set, and obtaining the error between the verification sample data sets to obtain a parameter prediction error;
and adjusting the parameters of the LSTM neural network by taking the parameter prediction error in a first preset range as a training target to obtain the data prediction model.
Optionally, the process of training the fault determination model includes:
dividing the target time interval into a plurality of sample time intervals according to the duration of the preset time interval;
obtaining the fault type of the sample equipment in each sample time interval;
inputting the sample data set in each sample time interval into a preset neural network;
respectively determining the output result of the preset neural network to the sample data sets in each sample time interval and the error between the sample data sets and the fault types corresponding to the sample time intervals to obtain fault prediction errors;
and adjusting the parameters of the preset neural network by taking the fault prediction error in a second preset range as a training target to obtain the fault prediction model.
Optionally, the device pictures of the plurality of target spectral bands include: a device picture in the infrared spectrum band, a device picture in the ultraviolet spectrum band, and a device picture in the visible spectrum band.
Optionally, the target parameters of the device picture in the infrared spectrum include: the hot spot temperature, the relative temperature difference and the average temperature of a preset position of the equipment;
the target parameters of the device picture in the ultraviolet spectral range include: discharge grade, discharge form and spot area;
the target parameters of the device picture in the visible spectrum band include: roughness, contrast and orientation.
Optionally, the method for predicting a device failure according to the first aspect of the present invention further includes:
and if the fault prediction result represents that the target equipment is abnormal, sending fault early warning information.
In a second aspect, the present invention provides an apparatus for predicting a device failure, including:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a device picture set of a target device at a plurality of sampling moments in a preset time interval;
wherein the device picture set comprises device pictures of a plurality of target spectral bands;
an extracting unit, configured to extract a reference value of a target parameter of each target spectral band device picture in each device picture set, to obtain a reference data set corresponding to each sampling time;
the data prediction unit is used for inputting each reference data set into a pre-trained data prediction model to obtain prediction data sets of the target equipment at a plurality of prediction moments;
wherein the prediction data set comprises a prediction value of each target parameter;
and the fault prediction unit is used for inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
In a third aspect, the present invention provides a server, comprising: a memory and a processor; the memory stores a program adapted to be executed by the processor to implement the method for predicting a failure of a device according to any one of the first aspect of the present invention.
Based on the above technical solution, in the device failure prediction method provided in the embodiments of the present invention, first, device picture sets of a target device at multiple sampling moments within a preset time interval are obtained, and any one of the device picture sets includes device pictures of multiple target spectral bands, then, a reference value of a target parameter of each target spectral band device picture in each device picture set is extracted, a reference data set corresponding to each sampling moment is obtained, each reference data set is further input into a pre-trained data prediction model, a prediction data set of the target device at multiple prediction moments is obtained, and finally, each prediction data set is input into a pre-trained failure determination model, so as to obtain a failure prediction result of the target device. The equipment fault prediction method provided by the invention can be used for predicting the fault of the target equipment based on the equipment pictures of different target spectra, further determining whether the power equipment is likely to have the fault in advance, providing a reference basis for operation and maintenance personnel to take targeted operation and maintenance measures and solving the problems in the prior art.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of an apparatus failure prediction method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a network architecture of an LSTM neural network according to an embodiment of the present invention;
fig. 3 is a block diagram of a device failure prediction apparatus according to an embodiment of the present invention;
fig. 4 is a block diagram of another device failure prediction apparatus according to an embodiment of the present invention;
fig. 5 is a block diagram of a server according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The equipment failure prediction method provided by the invention can be applied to electronic equipment, wherein the electronic equipment can be a notebook computer, a PC (personal computer) and other special failure prediction servers capable of acquiring picture information and operating preset failure prediction programs, and can also be applied to servers on a network side under certain conditions. Referring to fig. 1, a flow of the device failure prediction method provided in the embodiment of the present invention may include:
s100, acquiring a device picture set of a target device at a plurality of sampling moments in a preset time interval.
In practical applications, the power system includes a large number of electrical devices, such as insulators, transformers, high-voltage transmission lines, cables, disconnectors, high-voltage bushings, voltage-sharing rings, current transformers, voltage transformers, parallel capacitors, reactors, wave traps, switch cabinets, knife switches, and the like, and the target device in this step may be any one of the above electrical devices, and of course, may be any other related electrical device in the power system, which is not listed in the above description.
Optionally, after the target device is determined, a preset time interval may be obtained, and then a plurality of sampling moments are determined within the preset time interval according to a preset time interval, for example, the preset time interval is 8 in practical application: 00-9: 00, a preset time interval of 1 minute, 8: 00. 8: 01. 8: 02...9: 00 the time points are respectively taken as sampling time points, and the interval between the sampling time points is 1 minute. Of course, the specific selection of the specific preset time interval and the preset time interval needs to be selected in combination with the calculation power of the electronic device and the specific prediction precision requirement, and the specific values of the preset time interval and the preset time interval are not limited in the present invention.
It should be noted that, the fault determination model used in the subsequent fault prediction has strict requirements on the time range and the number of the input data and the time interval between each input data, and the preset time interval and the number of the sampling times mentioned in this step all should meet the requirements of the fault determination model, that is, the training process of the fault determination model needs to be selected in combination, and the specific training process will be developed in the subsequent content, which is not detailed here.
Further, after each sampling time is determined, the device picture sets of the target device at each sampling time can be obtained respectively. For any device picture set, device pictures of the target device in a plurality of target spectral bands are included.
Optionally, the device pictures of multiple target bands according to the embodiment of the present invention include: a device picture in the infrared spectrum band, a device picture in the ultraviolet spectrum band, and a device picture in the visible spectrum band. Of course, other device pictures in the spectrum may be included, and the present invention is also within the protection scope of the present invention without departing from the scope of the core idea of the present invention. It can be understood that, for any device picture set, the sampling time corresponding to each device picture in the same device picture set is the same, and the sampling time corresponding to different device picture sets is different.
It should be noted that, in a specific implementation manner of obtaining the device picture, the device picture set of the target device may be obtained at each sampling time according to the sampling time, or videos of the target device in different spectral bands within a preset time interval may be obtained, and then the device picture corresponding to each sampling time is extracted from the corresponding video, so that the device picture set may also be obtained.
S110, extracting reference values of target parameters of target spectrum device pictures in the device picture sets to obtain reference data sets corresponding to the sampling moments.
For each group of device picture sets obtained in S100, the reference value of the target parameter of each target spectral band device picture in the device picture set is extracted, and a corresponding reference data set is obtained.
The following describes a process of extracting a reference data set by taking any device picture set as an example:
firstly, extracting actual parameter values of target parameters of target spectrum device pictures in a device picture set to obtain an actual data set corresponding to each device picture set.
Specifically, the target parameters of the device pictures in different spectral bands may include:
device picture for infrared spectrum: the method comprises the steps of dividing target equipment and a background in an equipment picture of an infrared spectrum, extracting hot spot temperature and coordinates thereof from a preset position of the target equipment, and calculating relative temperature difference and average temperature of the preset position by combining normal temperature values of the preset position.
Device pictures for the ultraviolet spectrum: and (3) deducing the discharge grade, the discharge form and the light spot area of the target equipment by combining the ultraviolet and visible spectrum section image frames and using a fuzzy inference machine in the prior art.
For device pictures in the visible spectrum: mainly including roughness, contrast and orientation. For the roughness acquisition, the average intensity value of pixels in an active window with the size of 2k × 2k pixels in the device picture can be calculated; then, for each pixel, respectively calculating the average intensity difference between windows which are not overlapped in the horizontal direction and the vertical direction; the roughness is obtained by calculating the average of sbest over the whole image.
The contrast is obtained mainly by counting the intensity distribution of the pixels. And the direction degree needs to calculate all pixel gradient vectors, which are represented by a histogram, and the number of pixels larger than a given threshold value is calculated. The directionality of the image population can be obtained by calculating the sharpness of the peaks in the histogram. The specific acquisition process can be implemented based on the prior art, and is not described herein again.
It can be understood that, since the device picture sets correspond to the sampling time, the acquired actual parameter values of the target parameters of any device picture set also correspond to the sampling time corresponding to the device picture set.
Further, the actual data set of the device picture set is processed according to a preset preprocessing rule to obtain a corresponding preprocessed actual data set, wherein the preprocessed actual data set comprises preprocessed actual parameter values of the target parameters.
Based on the prior art, in the field of time series analysis, after an observed value sequence is obtained, the pure randomness and the stationarity of data are firstly checked, the process is called as sequence preprocessing, and the sequence can be divided into an irregular sequence, a stable sequence and a non-stable sequence according to the checking result. If the sequence is a non-stationary sequence, a difference operation is needed to convert the sequence into a differential stationary sequence. Based on this, this step also needs to perform certain preprocessing operations on each obtained actual data set to obtain a smooth type sequence meeting the use requirement, and as for the specific processing process, the implementation can be realized based on the prior art mentioned in the above, and the implementation is not expanded here.
After the preprocessing is finished, the preprocessed actual parameter values in the preprocessed actual data set are subjected to standardization processing, and finally the reference values of the target parameters of the target spectrum band device pictures in the device picture set are obtained.
Specifically, the normalization is to perform mean value removal and variance normalization on each feature dimension, so that the processed multi-spectral-band data conforms to the processing procedure of standard normal distribution, and specifically, the actual parameter values after each preprocessing can be normalized according to the following formula:
Figure BDA0003318994360000081
wherein x refers to the actual parameter value of the preprocessed target parameter, μ refers to the mean value of the target parameter, σ refers to the variance of the target parameter, and x refers to the reference value of the target parameter.
Of course, other normalization methods can be used to normalize the preprocessed actual parameter values, which are not listed here.
After the above normalization processing is performed on each target parameter, the obtained set is a reference data set, and correspondingly, all the device picture sets are traversed, so that the reference data set corresponding to each sampling moment can be obtained.
And S120, inputting each reference data set into a pre-trained data prediction model to obtain a prediction data set of the target equipment at a plurality of prediction moments.
Optionally, in order to predict the operating state of the target device, the present invention provides a data prediction model, where reference data sets of the target device at different sampling times obtained in the foregoing steps are input into the data prediction model, so that predicted data sets of the target device at multiple prediction times in a future period of time can be obtained, and any one of the predicted data sets further includes a predicted value of each target parameter.
It should be noted that, for the specific selection of the prediction duration and the number of the prediction moments, the specific setting and the specific prediction requirement of the data prediction model may be collected, and the input requirement setting of the fault determination model in the subsequent step, that is, the number of the prediction moments output by the data prediction model in this step, should satisfy the requirement on the number of the prediction data sets in S130.
As for the specific training process of the data prediction model, it will be mentioned in the following, and will not be described in detail here.
And S130, inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
The embodiment of the invention further provides a fault determination model, and the fault prediction result of the target equipment can be determined by inputting each prediction data set obtained by prediction in the previous step into the fault determination model. Optionally, the failure prediction result may be represented by a simple numerical value, for example, 0, 1, 2, and 3 represent the failure prediction result, where 0 represents that the target device does not fail in the time period corresponding to each prediction time, and 1 to 3 represent that the target device may have a corresponding type of failure in the time period corresponding to each prediction time.
Optionally, if the predicted failure prediction result represents that the target device is abnormal, for example, the prediction result is 1, failure early warning information is sent to notify the operation and maintenance personnel that the target device may fail, and special attention needs to be given in advance or corresponding protective measures need to be taken.
In summary, the device fault prediction method provided in the embodiments of the present invention can perform fault prediction on a target device based on device pictures of different multiple target spectral bands, realize prediction on time series data based on horizontal and vertical laws of a multi-dimensional time series between different dimensions and in time, and further determine whether a power device is likely to have a fault in advance by using a fault determination model according to a data prediction result, so as to provide a reference basis for operation and maintenance personnel to take targeted operation and maintenance measures, thereby solving the problems in the prior art.
Further, compared with the prior art, the equipment fault prediction method provided by the invention also has the following advantages:
1. operability. The power equipment fault early warning can be carried out without power failure maintenance, the cost is greatly reduced, and meanwhile, the operation risk of maintenance personnel is reduced. The power equipment fault early warning system can automatically operate, provides an early warning result which is easy to understand, and has small operation difficulty and low requirement on professional knowledge.
2. The comprehensiveness of the product. The method integrates the information of the infrared spectrum, the ultraviolet spectrum and the visible spectrum to comprehensively evaluate and predict the fault condition of the power equipment, is more comprehensive than a single spectrum fault maintenance scheme, and can provide fault information in advance.
3. And (4) accuracy. The method mainly adopts a deep learning method, learns the historical operating condition of the power equipment through a network, drives a training model by a large amount of data, and obtains a more accurate prediction result in a simulation data test stage.
The training process of the data prediction model and the fault determination model provided by the invention is introduced as follows:
the data prediction model provided by the embodiment of the invention is obtained based on LSTM neural network training. As can be seen from the network architecture diagram shown in fig. 2, the LSTM neural network includes an input layer, a hidden layer, and an output layer, where the input layer is used to input a training sample, set parameters such as seq _ length, batch _ size, and cell _ input, the hidden layer is used to learn a relationship between multi-spectral-segment data in a time domain and different feature dimensions, and the output layer outputs a predicted value of a target parameter.
As shown in the unit structure diagram of the LSTM neural network in fig. 2, the hidden layer of the long-short term memory network is used to transfer the long-term state and the short-term state between units, so as to learn the rules in the time domain by using the long-short term state, and finally achieve the purpose of prediction. Each unit is composed of a forgetting gate, an updating gate and an output gate.
Forget the door: as shown in fig. 2, the current state is represented by t, c(t-1)The long-term state at the previous moment is shown, and the door is forgotten to determine how much the long-term state at the previous moment is reserved.
And (4) updating the door: shown in FIG. 2 by x(t)The input at the current time is indicated, and the information content is updated to determine the input at the current time and retained.
An output gate: as shown in fig. 2, after the tanh activation function, the output gate determines the short-term state a at the current time(t)The information content.
Based on the basic architecture of the neural network, a sample picture set of a sample device at a plurality of sampling moments in a target time interval is obtained, wherein the sample picture set comprises a plurality of sample pictures of target spectral bands, and then sample values of target parameters of the sample pictures of the target spectral bands in each sample picture set are extracted to obtain the sample data set of each sampling moment.
It should be noted that the obtaining process of the sample data set may be implemented by referring to the implementation processes of S100 and S110 in the embodiment shown in fig. 1, and will not be repeated here. Different from the previous steps, in the training process of the data prediction model, the sample picture set not only comprises the picture set of the sample equipment in the normal operation state, but also comprises the picture set of the sample equipment in the fault state, so that a better training effect is achieved. In addition, in order to train the model as comprehensively as possible, the target time interval may be selected to be greater than the preset time interval, and generally, the target time interval may be divided into a plurality of preset time intervals with equal length.
After obtaining a plurality of sample data sets, dividing each sample data set into a training sample data set and a verification sample data set, optionally, further comprising a test sample data set, wherein the training sample data set is used for training an LSTM neural network, so that the model continuously fits the training set, the verification sample data set is mainly used for adjusting parameters of the neural network and primarily evaluating the model, and the test sample data set is used for verifying accuracy of an output result of the model.
Based on the classification of the sample data, the output result of the LSTM neural network to each training sample data set is respectively determined, the error between the LSTM neural network and the training sample data set is verified, a parameter prediction error is obtained, and then the parameter of the LSTM neural network is adjusted by taking the parameter prediction error in a first preset range as a training target until a data prediction model is obtained.
The first preset range is mainly determined based on the prediction precision requirement and the calculation force of the electronic equipment, and the specific setting of the first preset range is not limited by the invention.
Optionally, after the data prediction model is obtained, training of the fault determination model may be implemented based on the same sample picture set.
As mentioned above, the length of the aforementioned target time interval is greater than the duration of the preset time interval, and based on this, the target time interval is first divided into a plurality of sample time intervals according to the duration of the preset time interval, it can be understood that any sample time interval includes a plurality of sample data sets, and then the fault type of the sample device in each sample time interval is obtained.
Further, the sample data set in each sample time interval is input into a preset neural network, such as a full convolution network, similar to the data prediction model, and the preset neural network used for training the fault determination model also includes an input layer, a hidden layer and an output layer, where the input layer is used for inputting the sample data set and the fault type corresponding to each sample time interval, the hidden layer is used for learning the corresponding relationship between the sample data and the fault type, and the output layer is used for outputting the prediction result of the fault type. Based on the prior art, the main functional blocks of the hidden layer include a convolutional layer, a ReLU activation function, a BN normalization module, a pooling layer, and Softmax. The convolutional layer has a plurality of filters as a feature extraction layer. And carrying out batch normalization on each convolution layer, wherein the batch normalization is completed by a ReLU activation function. The BN layer forcibly pulls the distribution back to the standard normal distribution with the mean value of 0 and the variance of 1 through a normalization means, so that the gradient is increased, and the convergence rate of the model is improved. Global average pooling is used to reduce the number of parameters in the model prior to classification. Softmax was used for the final classification.
And respectively determining the output result of the preset neural network on the sample data sets in each sample time interval, obtaining a fault prediction error by the error between the fault types corresponding to the sample time intervals, and adjusting the parameters of the preset neural network by taking the fault prediction error in a second preset range as a training target until a fault prediction model is obtained.
It should be noted that the second preset range is mainly determined based on the prediction accuracy requirement and the calculation power of the electronic device, and the specific setting of the second preset range is not limited in the present invention.
The device failure prediction apparatus provided in the embodiment of the present invention is introduced below, and the device failure prediction apparatus described below may be regarded as a functional module architecture that needs to be set in the central device to implement the device failure prediction method provided in the embodiment of the present invention; the following description may be cross-referenced with the above.
Fig. 3 is a block diagram of a device failure prediction apparatus according to an embodiment of the present invention, and referring to fig. 3, the apparatus may include:
the device comprises an obtaining unit 10, a processing unit and a processing unit, wherein the obtaining unit is used for obtaining a device picture set of a target device at a plurality of sampling moments in a preset time interval;
wherein the device picture set comprises device pictures of a plurality of target spectral bands;
an extracting unit 20, configured to extract reference values of target parameters of target spectral band device pictures in each device picture set, so as to obtain a reference data set corresponding to each sampling time;
the data prediction unit 30 is configured to input each reference data set into a pre-trained data prediction model to obtain prediction data sets of the target device at multiple prediction moments;
the prediction data set comprises prediction values of all target parameters;
and the fault prediction unit 40 is used for inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
Optionally, the obtaining unit 10 is configured to obtain a device picture set of a target device at multiple sampling moments in a preset time interval, and includes:
acquiring a preset time interval;
determining a plurality of sampling moments in a preset time interval according to a preset time interval;
and respectively acquiring a device picture set of the target device at each sampling moment.
Optionally, the extracting unit 20 is configured to extract a reference value of a target parameter of each target spectral band device picture in each device picture set, and includes:
extracting actual parameter values of target parameters of target spectrum band device pictures in all device picture sets to obtain actual data sets corresponding to all the device picture sets;
processing each actual data set according to a preset preprocessing rule to obtain a corresponding preprocessed actual data set;
the preprocessed actual data set comprises preprocessed actual parameter values of all target parameters;
and respectively carrying out standardization processing on each preprocessed actual parameter value in each preprocessed actual data set to obtain a reference value of a target parameter of each target spectrum device picture in each device picture set.
Optionally, referring to fig. 4, fig. 4 is a block diagram of another device failure prediction apparatus provided in the embodiment of the present invention, and on the basis of the embodiment shown in fig. 3, the failure prediction apparatus provided in this embodiment further includes:
and the alarm unit 50 is used for sending fault early warning information if the fault prediction result represents that the target equipment is abnormal.
Fig. 5 is a hardware structure diagram of a server according to an embodiment of the present invention, which is shown in fig. 5 and may include: at least one processor 100, at least one communication interface 200, at least one memory 300, and at least one communication bus 400;
in the embodiment of the present invention, the number of the processor 100, the communication interface 200, the memory 300, and the communication bus 400 is at least one, and the processor 100, the communication interface 200, and the memory 300 complete the communication with each other through the communication bus 400; it is clear that the communication connections shown by the processor 100, the communication interface 200, the memory 300 and the communication bus 400 shown in fig. 5 are merely optional;
optionally, the communication interface 200 may be an interface of a communication module, such as an interface of a GSM module;
the processor 100 may be a central processing unit CPU or an application Specific Integrated circuit asic or one or more Integrated circuits configured to implement embodiments of the present invention.
Memory 300 may comprise high-speed RAM memory and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
The processor 100 is specifically configured to execute an application program in the memory to implement the steps of the device failure prediction method described above.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. An apparatus failure prediction method, comprising:
acquiring a device picture set of a target device at a plurality of sampling moments within a preset time interval;
wherein the device picture set comprises device pictures of a plurality of target spectral bands;
extracting reference values of target parameters of target spectrum device pictures in the device picture sets to obtain reference data sets corresponding to the sampling moments;
inputting each reference data set into a pre-trained data prediction model to obtain prediction data sets of the target equipment at a plurality of prediction moments;
wherein the prediction data set comprises a prediction value of each target parameter;
and inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
2. The method according to claim 1, wherein the obtaining a device picture set of a target device at a plurality of sampling moments within a preset time interval comprises:
acquiring a preset time interval;
determining a plurality of sampling moments in a preset time interval according to the preset time interval;
and respectively acquiring a device picture set of the target device at each sampling moment.
3. The method according to claim 1, wherein the extracting a reference value of a target parameter of each target spectral band device picture in each device picture set comprises:
extracting actual parameter values of target parameters of target spectrum device pictures in the device picture sets to obtain actual data sets corresponding to the device picture sets;
processing each actual data set according to a preset preprocessing rule to obtain a corresponding preprocessed actual data set;
the preprocessed actual data set comprises preprocessed actual parameter values of the target parameters;
and respectively carrying out standardization processing on each preprocessed actual parameter value in each preprocessed actual data set to obtain a reference value of a target parameter of each target spectrum device picture in each device picture set.
4. The equipment failure prediction method of claim 1, wherein the process of training the data prediction model comprises:
acquiring a sample picture set of a plurality of sampling moments of sample equipment in a target time interval;
wherein the sample picture set comprises sample pictures of a plurality of target spectral bands;
extracting sample values of target parameters of each target spectrum band sample picture in each sample picture set to obtain a sample data set of each sampling moment;
dividing each sample data set into a training sample data set and a verification sample data set;
respectively determining the output result of the LSTM neural network to each training sample data set, and obtaining the error between the verification sample data sets to obtain a parameter prediction error;
and adjusting the parameters of the LSTM neural network by taking the parameter prediction error in a first preset range as a training target to obtain the data prediction model.
5. The device failure prediction method of claim 4, wherein the process of training the failure determination model comprises:
dividing the target time interval into a plurality of sample time intervals according to the duration of the preset time interval;
obtaining the fault type of the sample equipment in each sample time interval;
inputting the sample data set in each sample time interval into a preset neural network;
respectively determining the output result of the preset neural network to the sample data sets in each sample time interval and the error between the sample data sets and the fault types corresponding to the sample time intervals to obtain fault prediction errors;
and adjusting the parameters of the preset neural network by taking the fault prediction error in a second preset range as a training target to obtain the fault prediction model.
6. The device failure prediction method of claim 1, wherein the device pictures of the plurality of target spectral bands comprise: a device picture in the infrared spectrum band, a device picture in the ultraviolet spectrum band, and a device picture in the visible spectrum band.
7. The device failure prediction method of claim 6, wherein the target parameters of the device picture in the infrared spectrum comprise: the hot spot temperature, the relative temperature difference and the average temperature of a preset position of the equipment;
the target parameters of the device picture in the ultraviolet spectral range include: discharge grade, discharge form and spot area;
the target parameters of the device picture in the visible spectrum band include: roughness, contrast and orientation.
8. The device failure prediction method of any one of claims 1-7, further comprising:
and if the fault prediction result represents that the target equipment is abnormal, sending fault early warning information.
9. An apparatus for predicting a failure of a device, comprising:
the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a device picture set of a target device at a plurality of sampling moments in a preset time interval;
wherein the device picture set comprises device pictures of a plurality of target spectral bands;
an extracting unit, configured to extract a reference value of a target parameter of each target spectral band device picture in each device picture set, to obtain a reference data set corresponding to each sampling time;
the data prediction unit is used for inputting each reference data set into a pre-trained data prediction model to obtain prediction data sets of the target equipment at a plurality of prediction moments;
wherein the prediction data set comprises a prediction value of each target parameter;
and the fault prediction unit is used for inputting each prediction data set into a pre-trained fault determination model to obtain a fault prediction result of the target equipment.
10. A server, comprising: a memory and a processor; the memory stores a program adapted to be executed by the processor to implement the device failure prediction method of any one of claims 1 to 8.
CN202111240270.6A 2021-10-25 2021-10-25 Equipment fault prediction method and device and server Pending CN113971777A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111240270.6A CN113971777A (en) 2021-10-25 2021-10-25 Equipment fault prediction method and device and server

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111240270.6A CN113971777A (en) 2021-10-25 2021-10-25 Equipment fault prediction method and device and server

Publications (1)

Publication Number Publication Date
CN113971777A true CN113971777A (en) 2022-01-25

Family

ID=79588191

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111240270.6A Pending CN113971777A (en) 2021-10-25 2021-10-25 Equipment fault prediction method and device and server

Country Status (1)

Country Link
CN (1) CN113971777A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116705271A (en) * 2023-08-09 2023-09-05 山东博达医疗用品股份有限公司 Big data medical treatment flushing equipment operation monitoring system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116705271A (en) * 2023-08-09 2023-09-05 山东博达医疗用品股份有限公司 Big data medical treatment flushing equipment operation monitoring system
CN116705271B (en) * 2023-08-09 2023-11-14 山东博达医疗用品股份有限公司 Medical flushing equipment operation monitoring system based on big data

Similar Documents

Publication Publication Date Title
James et al. Intelligent time-adaptive transient stability assessment system
CN109544399B (en) Power transmission equipment state evaluation method and device based on multi-source heterogeneous data
CN110829417A (en) Electric power system transient stability prediction method based on LSTM double-structure model
CN107436277A (en) The single index data quality control method differentiated based on similarity distance
CN106779215B (en) Machine learning-based power grid global delay situation sensing method
CN114298175A (en) Power equipment state monitoring and fault early warning method and system based on edge calculation
Cortes-Robles et al. Fast-training feedforward neural network for multi-scale power quality monitoring in power systems with distributed generation sources
CN111160626A (en) Power load time sequence control method based on decomposition and fusion
CN111506635A (en) System and method for analyzing residential electricity consumption behavior based on autoregressive naive Bayes algorithm
KR102315580B1 (en) Fire predictive analysis device and method of building
CN116610998A (en) Switch cabinet fault diagnosis method and system based on multi-mode data fusion
CN113971777A (en) Equipment fault prediction method and device and server
Jafarzadeh et al. Real-time transient stability prediction of power systems based on the energy of signals obtained from PMUs
Gjorgiev et al. Simulation-driven deep learning for locating faulty insulators in a power line
CN112949201B (en) Wind speed prediction method and device, electronic equipment and storage medium
US11906957B1 (en) Method for assessing short-term voltage stability of power grid based on synchronous spatio-temporal information learning
Ren et al. Online anomaly detection using machine learning and HPC for power system synchrophasor measurements
CN112419243A (en) Power distribution room equipment fault identification method based on infrared image analysis
Srikanth et al. A novel three-dimensional deep learning algorithm for classification of power system faults
Omran et al. Utilizing different types of deep learning models for classification of series arc in photovoltaics systems
CN111506636A (en) System and method for analyzing residential electricity consumption behavior based on autoregressive and neighbor algorithm
CN116485135A (en) Wisdom commercial building big data management platform
CN114566964B (en) Power distribution network feeder automation control method, device, equipment and storage medium
CN107147143B (en) Method for establishing early warning model of fan interlocking off-line fault
CN113361730B (en) Risk early warning method, device, equipment and medium for maintenance plan

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