CN110334756A - Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium - Google Patents
Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium Download PDFInfo
- Publication number
- CN110334756A CN110334756A CN201910563871.7A CN201910563871A CN110334756A CN 110334756 A CN110334756 A CN 110334756A CN 201910563871 A CN201910563871 A CN 201910563871A CN 110334756 A CN110334756 A CN 110334756A
- Authority
- CN
- China
- Prior art keywords
- warning information
- power system
- system monitor
- alarm event
- sample set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000009434 installation Methods 0.000 title description 2
- 238000012549 training Methods 0.000 claims abstract description 96
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 45
- 238000012544 monitoring process Methods 0.000 claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims description 26
- 230000015572 biosynthetic process Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 3
- 238000011161 development Methods 0.000 abstract description 5
- 230000005856 abnormality Effects 0.000 abstract description 4
- 238000012545 processing Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 230000018109 developmental process Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000005194 fractionation Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Economics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Human Resources & Organizations (AREA)
- Public Health (AREA)
- Power Engineering (AREA)
- Water Supply & Treatment (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Alarm Systems (AREA)
Abstract
The invention discloses method, apparatus, terminal device and the media of a kind of identification of power system monitor alarm event.This method comprises: obtaining the warning information of power system monitor；Alarm event is tentatively identified according to warning information, forms preliminary identification data；To warning information and preliminary identification data prediction, training sample set is formed；Training sample set includes the event of multiple definition；Using training sample set to convolutional neural networks model training, power system monitor alarm event identification model is formed；Monitoring alarm event is identified using power system monitor alarm event identification model.Improve efficiency and accuracy that power system monitor warning information is identified as alarm event, effectively alleviate the prison screen ability of power system monitor business personnel, the working efficiency of the daily monitoring of business personnel and accident abnormality disposition is improved, so that the development trend that geometry grade increases is presented in the quantity that power system monitor warning information platform preferably adapts to the equipment fault warning information of power grid.
Description
Technical field
The present embodiments relate to electric power supervisory control technical fields more particularly to a kind of power system monitor alarm event to identify
Method, apparatus, terminal device and medium.
Background technique
As grid equipment popularization and intellectual monitoring level are promoted, geometry is presented in the quantity of equipment fault warning information
Grade growth trend, existing power system monitor business rely on the surveillance style that warning information responds one by one, need to each information by
One is differentiated, analyzed and is made feedback, since the monitoring device quantity being related to is various, monitoring information amount is big, abnormal failure feelings
Condition is complicated, only according to the technological means such as traditional warning information window, alarm window list are borrowed, is easy to appear equipment fault or exception is failed to judge
The case where erroneous judgement, causes biggish prison to monitoring personnel and shields pressure, and can not adapt to power system monitor industry under existing situation
The requirements at the higher level of business.
Summary of the invention
The present invention provides method, apparatus, terminal device and the medium of a kind of power system monitor alarm event identification, improves monitoring
Warning information identifies the efficiency and accuracy of alarm event, preferably adapts to the requirement of power system monitor business under existing situation.
In a first aspect, the embodiment of the invention provides a kind of power system monitor alarm events to know method for distinguishing, comprising:
Obtain the warning information of power system monitor；
Alarm event is tentatively identified according to the warning information, forms preliminary identification data；
To the warning information and the preliminary identification data prediction, training sample set is formed；The training sample set
Event including multiple definition；
Using the training sample set to convolutional neural networks model training, forms power system monitor alarm event and identify mould
Type；
Monitoring alarm event is identified using the power system monitor alarm event identification model.
Optionally, the warning information includes power system monitor alarm event historical information, substation and line name statistics
Metric data when information and event occur.
Optionally, described that alarm event is tentatively identified according to the warning information, form preliminary identification data, comprising:
The warning information is tentatively identified into alarm event by rule base, forms preliminary identification data.
Optionally, described that the warning information is wrapped with the preliminary identification data prediction, formation training sample set
It includes:
The warning information and the preliminary identification data are segmented and stop words is gone to handle；
Will treated warning information vectorization, form training sample set；Wherein, any one event of the training sample set
Warning information be data matrix, the most major term length of the line number of the data matrix is that treated warning information, the number
It is the dimension of treated warning information vectorization according to matrix column number.
Optionally, it is described will treated warning information vectorization, form training sample set, comprising:
By treated, warning information passes through word2vec formation term vector, and the term vector constitutes the training sample
Collection.
Optionally, described to use the training sample set to convolutional neural networks model training, form power system monitor alarm
Event recognition model, comprising:
The warning information of input is normalized using the input layer of the convolutional neural networks model；
Using multiple convolution units in the convolutional layer of the convolutional neural networks model to normalized warning information into
Row convolution forms convolution characteristic pattern；
The convolution characteristic pattern maximum pondization is operated using the pond layer of the convolutional neural networks model, is formed maximum
Value Data matrix；
The full articulamentum of the convolutional neural networks model is used to calculate the warning information of the input as each definition
The probability of event, and the recognition result of the event of maximum probability as the warning information of the input is chosen, it completes to the volume
Product neural network model training, forms power system monitor alarm event identification model.
Second aspect, the embodiment of the invention also provides a kind of devices of power system monitor alarm event identification, comprising:
The warning information of power system monitor obtains module, for obtaining the warning information of power system monitor；
Preliminary identification data form module, for tentatively identifying monitoring alarm event according to the warning information, are formed just
Step identification data；
Training sample set forms module, for forming instruction to the warning information and the preliminary identification data prediction
Practice sample set；The training sample set includes the event of multiple definition；
Power system monitor alarm event identification model forms module, for using the training sample set to convolutional neural networks
Model training forms power system monitor alarm event identification model；
Power system monitor alarm event identification module, for using power system monitor alarm event identification model identification monitoring
Alarm event.
Optionally, it includes regular library unit that the preliminary identification data, which form module,；
The rule library unit is used to the warning information tentatively identifying monitoring alarm event by rule base, be formed just
Step identification data.
The third aspect, the embodiment of the invention also provides a kind of terminal devices, comprising:
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes that the power system monitor alarm event that any embodiment of that present invention provides knows method for distinguishing.
The third aspect, the embodiment of the invention also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, which is characterized in that the power system monitor for realizing that any embodiment of that present invention provides when the program is executed by processor alerts thing
Part knows method for distinguishing.
The technical solution of the embodiment of the present invention tentatively identifies alarm event to the warning information of acquisition, forms preliminary identification
After data, warning information and preliminary identification data prediction are formed into training sample set, improve the validity of training sample set
And accuracy.And training sample set is used to form power system monitor alarm event identification model to convolutional neural networks model training,
Improve the classification high efficiency and accuracy of power system monitor alarm event identification model.Therefore, power grid prison provided in this embodiment
The warning information autonomous classification to power system monitor may be implemented in the method for charging alert event recognition, improves power system monitor alarm letter
Breath is identified as the efficiency and accuracy of alarm event, effectively alleviates the prison screen ability of power system monitor business personnel, improves
The working efficiency of the daily monitoring of business personnel and accident abnormality disposition, so that power system monitor warning information platform preferably adapts to electricity
The development trend that geometry grade increases is presented in the quantity of the equipment fault warning information of net.
Detailed description of the invention
Fig. 1 is the flow chart that a kind of power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing；
Fig. 2 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing；
Fig. 3 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing；
Fig. 4 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing；
Fig. 5 is a kind of schematic diagram of convolutional neural networks structure provided in an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of the device of power system monitor alarm event identification provided in an embodiment of the present invention.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
The embodiment of the invention provides a kind of power system monitor alarm event identification method, apparatus, equipment and storage medium,
The warning information autonomous classification to power system monitor may be implemented, improve the effect that power system monitor warning information is identified as alarm event
Rate and accuracy effectively alleviate the prison screen ability of power system monitor business personnel, improve the daily monitoring of business personnel and thing
Therefore the working efficiency disposed extremely, so that power system monitor warning information platform preferably adapts to the equipment fault warning information of power grid
Quantity present geometry grade increase development trend.
Fig. 1 is the flow chart that a kind of power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing.This implementation
Example is applicable to the case where identifying alarm event by the warning information of power system monitor, and this method can alert thing by power system monitor
Part identification device executes, which can integrate on processor.As shown in Figure 1, this method specifically comprises the following steps:
S110, the warning information for obtaining power system monitor.
Specifically, power system monitor can be a monitor supervision platform, for monitoring the operation conditions of grid equipment.Ordinary circumstance
Under, power system monitor can be monitored the operating condition of the grid equipment in a city as unit of city.It is run in grid equipment
During, if there is alarm event, then the information that will characterize grid equipment alarm event is transmitted to power system monitor, makees
For warning information.The warning information of one alarm event generally comprises power system monitor alarm event historical information, substation and line
Metric data etc. when road title statistical information and event occur.Wherein, power system monitor alarm event historical information, including necessary announcement
Alert information and association warning information；Necessary warning information, including the total action message of protection act information, accident and protection are overlapped
Lock action message；Association warning information, including the movement of switch control circuit broken string, switch control circuit broken string involution and switch
Spring unstored energy movement etc..
S120, alarm event is tentatively identified according to warning information, form preliminary identification data.
Specifically, tentatively identification data are that the recognition result data of preliminary identification alarm event are carried out to warning information.One
As in the case of, it is preliminary to identify that data reinforce contacting between warning information and alarm event, so as to by tentatively knowing
Other data improve the validity and accuracy of training sample set, and then improve the power grid prison formed using training sample set training
Charge the classification high efficiency and accuracy of alert event recognition model.In addition, alarm event is classified generally according to grid equipment.
Illustratively, alarm event may include the alarm event of route, the alarm event of transformer, capacitive reactance device alarm event and connect
The alarm event on ground.And the alarm event of route can also include the alarm event of bus, the alarm event of transformer can be with
The alarm event of alarm event and time change including main transformer.The warning information of different alarm events is different.
S130, it to warning information and tentatively identifies data prediction, forms training sample set；Training sample set includes multiple
The event of definition.
Specifically, training sample set includes pretreated warning information and preliminary identification data.Warning information is alarm
The raw information of event includes all information of alarm event.Preliminary identification data are tentatively identified to warning information
The recognition result of alarm event.It is formed simultaneously training sample set by warning information and preliminary identification data, training can be improved
The validity and accuracy of sample set, and then improve the power system monitor alarm event identification formed using training sample set training
The classification high efficiency and accuracy of model.
Training sample concentrates the event for the multiple definition for including that can be defined according to the classification of alarm event.One thing
The warning information of an alarm event can occur for a grid equipment and tentatively identify data for part.
S140, using training sample set to convolutional neural networks model training, form power system monitor alarm event and identify mould
Type.
Specifically, convolutional neural networks model is a kind of deep learning algorithm.Deep learning algorithm is using successively training
Mode handles data, can obtain the advanced features expression for being originally inputted variables collection, improves prediction and classification is accurate
Property, it is widely used in characteristic processing problem and big data scene.Currently, the common model of deep learning algorithm has limited glass
The graceful machine of Wurz (Restricted Boltzmann Machine, RBM), self-encoding encoder (Auto Encoder, AE), deep layer conviction
Network (Deep Belief Network, DBN) and convolutional neural networks (Convolutional Neural Network, CNN)
Deng.And convolutional neural networks model has the characteristics that part connects and weight is shared.Part connection can reduce neural network
Nuisance parameter can design the more neural network models of the number of plies.Weight is shared to enable the feature in input data fine
Ground is identified, without being influenced by its present position.The expansibility of convolutional neural networks structure is very strong, it can use very deep
The number of plies, be capable of handling more complicated classification problem.Meanwhile the historical data of power system monitor is much Boolean type data, and roll up
Product neural network has good 0,1 variable processing capacity, therefore forms power system monitor alarm using convolutional neural networks model
The data-handling capacity of power system monitor alarm event identification model and the standard of warning information identification can be improved in event recognition model
True property.
S150, monitoring alarm event is identified using power system monitor alarm event identification model.
Specifically, after forming power system monitor alarm event identification model, power system monitor alarm event identification model is to defeated
The warning information entered carries out identification judgement, and warning information is categorized into different alarm events.
The technical solution of the present embodiment tentatively identifies alarm event to the warning information of acquisition, forms preliminary identification data
Afterwards, warning information and preliminary identification data prediction are formed into training sample set, improves the validity and standard of training sample set
Exactness.And training sample set is used to form power system monitor alarm event identification model to convolutional neural networks model training, it improves
The classification high efficiency and accuracy of power system monitor alarm event identification model.Therefore, power system monitor provided in this embodiment is accused
The warning information autonomous classification to power system monitor may be implemented in the method for alert event recognition, improves the knowledge of power system monitor warning information
Not Wei alarm event efficiency and accuracy, effectively alleviate the prison screen ability of power system monitor business personnel, improve business
The working efficiency of the daily monitoring of personnel and accident abnormality disposition, so that power system monitor warning information platform preferably adapts to power grid
The development trend that geometry grade increases is presented in the quantity of equipment fault warning information.
Fig. 2 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing.
As shown in Fig. 2, this method comprises:
S210, the warning information for obtaining power system monitor.
S220, warning information is tentatively identified into alarm event by rule base, forms preliminary identification data.
Specifically, rule base is that the logic that regulation person judges processing warning information is simulated according to warning information, and formulation is led to
The rule for crossing warning information identification alarm event is integrated.Therefore, rule base can also be known as Expert Rules library.Rule base can wrap
Include the logical relation and topological relation and the grade of the alarm event of identification etc. between different warning information features.Pass through rule
Then library tentatively identifies alarm event, can be similar to the logic judgment of regulation person, includes pretreatment by training sample set therefore
Preliminary identification data afterwards, can be improved the validity and accuracy of training sample set.
S230, it to warning information and tentatively identifies data prediction, forms training sample set；Training sample set includes multiple
The event of definition.
S240, using training sample set to convolutional neural networks model training, form power system monitor alarm event and identify mould
Type.
S250, monitoring alarm event is identified using power system monitor alarm event identification model.
Warning information is tentatively identified alarm event by rule base, forms preliminary identification by the technical solution of the present embodiment
Data.Using preliminary identification data as a part of training sample set, the validity and accuracy of training sample set can be improved.
And training sample set is used to form power system monitor alarm event identification model to convolutional neural networks model training, improve power grid
The classification high efficiency and accuracy of monitoring alarm event recognition model.
Fig. 3 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing.
As shown in figure 3, this method comprises:
S310, the warning information for obtaining power system monitor.
S320, alarm event is tentatively identified according to warning information, form preliminary identification data.
S331, warning information and preliminary identification data are segmented and stop words is gone to handle.
Specifically, warning information is generally text information, in order to realize the identification to warning information, can first believe alarm
Breath carries out word segmentation processing, i.e., carries out word fractionation to warning information, so that warning information forms multiple words, and counts alarm letter
All words in breath.After word segmentation processing, the stop words in warning information is rejected using deactivated vocabulary.Specifically, stop words
Table is the set of some words for being unable to characterize warning information, such as " ", the words such as " " and " ".Vocabulary is deactivated establishing
When, it can be established according to the electric power dictionary of profession, the word that will be unable to characterization warning information feature is integrated as stop words
To deactivate vocabulary.In addition, further include the specialized word of power grid in electric power dictionary, when carrying out word segmentation processing to warning information, electricity
Specialized word in power dictionary is segmented as a word.For example, disconnecting switch is the specialized word in electric power dictionary,
It is segmented when participle as a word.
S332, will treated warning information vectorization, form training sample set；Wherein, any thing of training sample set
The warning information of part is data matrix, the most major term length of the line number of data matrix is that treated warning information, data matrix
Columns be treated warning information vectorization dimension.
Specifically, treated warning information by word2vec can be formed into term vector, the term vector composition instruction
Practice sample set.Word2vec is the correlation model that a group is used to generate term vector.These models are the neural network of shallow-layer bilayer,
For training with the word text of construction linguistics again.Term vector forms training sample set, a thing in the form of data matrix
The warning information of part can be a data matrix.Illustratively, the data matrix A that training sample is concentrated is the announcement of an event
The data matrix that alert information is formed, specifically:
Wherein, n is the most major term length of warning information, and k is the dimension of warning information vectorization.
After carrying out participle and stop words processing to warning information, the training sample set of formation is particularly suited for training electricity
Net monitoring alarm event recognition model.
S340, using training sample set to convolutional neural networks model training, form power system monitor alarm event and identify mould
Type.
S350, monitoring alarm event is identified using power system monitor alarm event identification model.
Fig. 4 is the flow chart that another power system monitor alarm event provided in an embodiment of the present invention knows method for distinguishing.
As shown in figure 4, this method comprises:
S410, the warning information for obtaining power system monitor.
S420, alarm event is tentatively identified according to warning information, form preliminary identification data.
S430, it to warning information and tentatively identifies data prediction, forms training sample set；Training sample set includes multiple
The event of definition.
S441, the warning information of input is normalized using the input layer of convolutional neural networks model.
Specifically, Fig. 5 is a kind of schematic diagram of convolutional neural networks structure provided in an embodiment of the present invention.As shown in figure 5,
Input layer input is data matrix by pretreated warning information.Input layer carries out normalizing to the data matrix of warning information
Change, the feature vector of available warning information, for reflecting that the word in warning information is important in this warning information
Degree and weighted value.
S442, using multiple convolution units in the convolutional layer of convolutional neural networks model to normalized warning information into
Row convolution forms convolution characteristic pattern.
Specifically, convolutional layer includes multiple convolution units, and each convolution unit is different Feature Mapping.Each feature is reflected
Multiple mutually disjunct independent neuron, each neurons have been penetrated to be connected with the local receptor field of input layer.To input layer
The normalized data matrix of output carries out convolution operation, then exports feature are as follows:
ci=f (Ai:i+h-1×Wi+bi)
Wherein, WiIndicating i-th of convolution unit, the width l of convolution unit and the dimension k of warning information vectorization are consistent,
And the height h of convolution unit is then related with the range for the local warning information text feature to be extracted；ciFor i-th of convolution unit
The feature of output；Ai:i+h-1For the i-th row in the warning information data matrix of input to (i+h-1) row；biFor bias term；f
() is nonlinear activation function.
In order to accelerate convergence speed to the maximum extent, select ReLU function as nonlinear activation function, are as follows:
ReLU (x)=max (0, x)
After convolution effect by multiple convolution units of convolutional layer, the convolution characteristic pattern of generation are as follows:
C={ c1,c2,c3…cn-h+1}
Wherein, n is the most major term length of warning information, and h is the height of convolution unit.
Compared with traditional multilayer perceptron, in convolutional neural networks the weight of convolutional layer it is shared make it is trainable in network
Parameter tails off, and reduces network model complexity, over-fitting is reduced, to obtain a better generalization ability.
S443, convolution characteristic pattern maximum pondization is operated using the pond layer of convolutional neural networks model, forms maximum value
Data matrix.
Specifically, in order to find most representational local optimum output feature from every convolution characteristic pattern,
The convolution characteristic pattern for needing to extract convolutional layer carries out maximum pondization operation, to complete the data matrix from warning information
It is middle to concentrate all processes for extracting some local feature.The function in maximum pond are as follows:
vi=poolingmax(ci)
Wherein, viFor maximum pond function poolingmaxAct on output feature ciThe maximum value of generation.
The operation of maximum pondization is a nonlinear sub-sample function, for generate a maximum value sub-sample square
Battle array, further reduces the vector dimension of the corresponding data matrix of warning information, plays the effect that parameter about subtracts, be further reduced
The number of parameters relied on when final classification.
For m convolution unit, m feature vector of generation can be indicated are as follows:
Vk={ v1,v2,v3…vm}
After the operation of convolutional layer and pond layer, the warning information data matrix A of input meant that at feature to
Measure Vk。
S444, the full articulamentum of convolutional neural networks model is used to calculate the warning information of input as the event of each definition
Probability, and the recognition result for the event warning information as input for choosing maximum probability.
Specifically, it is each that full articulamentum, which can be used soft max function model and calculate the warning information of any input,
The probability of definition event chooses recognition result of the event as the warning information of maximum probability, completes to convolutional neural networks
Model training forms power system monitor alarm event identification model.The input sample collection of full articulamentum are as follows:
{(x(1),y(1))(x(2),y(2))(x(2),y(2))…(x(m),y(m))}
In formula, input feature vector x(i)For the feature vector of warning information, y(i)For the classification of event, m is of convolution unit
Number.
Then warning information x(l)The probability for being identified as event k is
In formula, W(k)For weight corresponding with event k, b(k)For bias term corresponding with event k, P (k) is warning information x(l)Correspond to the probability of event k.
In addition, Dropout strategy can be introduced when layer parameter is trained for connecting entirely, i.e., Shi Douhui is being updated every time
A part of trained parameter is randomly choosed to be given up.
S450, monitoring alarm event is identified using power system monitor alarm event identification model.
In addition, back-propagation algorithm (Back Propagation can be used in training convolutional neural networks model
Algorithm) be trained, i.e., so that model objective function is converged to minimum by constantly iteration, obtain optimal weight and
Bias term.In addition, there is over-fitting in training process in order to prevent, come using L2 regularization to convolutional neural networks
Parameter is constrained.
The embodiment of the present invention also provides a kind of device of power system monitor alarm event identification.Fig. 6 mentions for the embodiment of the present invention
A kind of structural schematic diagram of the device of the power system monitor alarm event identification supplied.As shown in fig. 6, the device includes:
The warning information of power system monitor obtains module 10, for obtaining the warning information of power system monitor.
Specifically, warning information may include power system monitor alarm event historical information, substation and line name statistics
Metric data when information and event occur.
Preliminary identification data form module 20, for tentatively identifying monitoring alarm event according to warning information, are formed preliminary
Identify data.
Training sample set forms module 30, for forming training sample to warning information and preliminary identification data prediction
Collection, training sample set includes the event of multiple definition.
Specifically, training sample set forms module 30 and can be segmented and be gone to stop to warning information and preliminary identification data
Word processing, and will treated warning information vectorization, form training sample set.Wherein, any one event of training sample set
Warning information be data matrix, the most major term length of the line number of data matrix is that treated warning information, data matrix
Columns is the dimension of treated warning information vectorization.Furthermore it is possible to by treated, warning information passes through word2vec shape
At term vector, term vector constitutes the training sample set.
Power system monitor alarm event identification model forms module 40, for using training sample set to convolutional neural networks mould
Type training forms power system monitor alarm event identification model.
Specifically, power system monitor alarm event identification model forms module 40 using the input layer of convolutional neural networks model
Warning information normalization to input, then using multiple convolution units in the convolutional layer of convolutional neural networks model to normalizing
The warning information of change carries out convolution, forms convolution characteristic pattern, later using the pond layer of convolutional neural networks model to convolution spy
Sign schemes maximum pondization operation, forms maximum value data matrix, is finally calculated using the full articulamentum of convolutional neural networks model defeated
The warning information entered is the probability of the event of each definition, and the knowledge for the event warning information as input for choosing maximum probability
Not as a result, completing to form power system monitor alarm event identification model to convolutional neural networks model training.
Power system monitor alarm event identification module 50, for being accused using the identification monitoring of power system monitor alarm event identification model
Alert event.
The technical solution of the present embodiment tentatively identifies alarm thing to warning information by tentatively identifying that data form module
Part, after forming preliminary identification data, training sample set forms module and warning information and preliminary identification data prediction is formed instruction
Practice sample set, improves the validity and accuracy of training sample set.And power system monitor alarm event identification model forms module
Using training sample set to convolutional neural networks model training, power system monitor alarm event identification model is formed, power grid is improved
The classification high efficiency and accuracy of monitoring alarm event recognition model.Therefore, power system monitor alarm event provided in this embodiment
Knowing method for distinguishing may be implemented the warning information autonomous classification to power system monitor, improves power system monitor warning information and is identified as accusing
The efficiency and accuracy of alert event, effectively alleviate the prison screen ability of power system monitor business personnel, improve business personnel day
The working efficiency of often monitoring and accident abnormality disposition, so that power system monitor warning information platform preferably adapts to the equipment event of power grid
The development trend that geometry grade increases is presented in the quantity for hindering warning information.
It is preliminary to identify that data form module 20 and include regular library unit 21 with continued reference to Fig. 6.
Regular library unit 21 is used to warning information tentatively identifying monitoring alarm event by rule base, is formed and tentatively identified
Data.
The embodiment of the present invention also provides a kind of terminal device.The equipment includes:
One or more processors；
Memory, for storing one or more programs；
When one or more programs are executed by one or more processors, so that one or more processors realize the present invention
The power system monitor alarm event that any embodiment provides knows method for distinguishing.It is mentioned because the equipment can execute any embodiment of that present invention
The power system monitor alarm event of confession knows method for distinguishing, therefore the power system monitor alarm event provided with any embodiment of that present invention
Know the beneficial effect of method for distinguishing, details are not described herein again.
The embodiment of the present invention also provides a kind of computer readable storage medium, is stored thereon with computer program, the program
The power system monitor alarm event for realizing that any embodiment of that present invention provides when being executed by processor knows method for distinguishing.Illustratively,
Storage medium can be floppy disk, the read-only memory (Read-Only Memory, ROM), random access memory of computer
(Random Access Memory, RAM), flash memory (FLASH), hard disk or CD etc., because of computer readable storage medium, this can
Method for distinguishing is known to execute the power system monitor alarm event of any embodiment of that present invention offer, therefore has the present invention is any to implement
The power system monitor alarm event that example provides knows the beneficial effect of method for distinguishing, and details are not described herein again.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (10)
1. a kind of power system monitor alarm event knows method for distinguishing characterized by comprising
Obtain the warning information of power system monitor；
Alarm event is tentatively identified according to the warning information, forms preliminary identification data；
To the warning information and the preliminary identification data prediction, training sample set is formed；The training sample set includes
The event of multiple definition；
Using the training sample set to convolutional neural networks model training, power system monitor alarm event identification model is formed；
Monitoring alarm event is identified using the power system monitor alarm event identification model.
2. the method according to claim 1, wherein the warning information includes power system monitor alarm event history
Metric data when information, substation and line name statistical information and event occur.
3. the method according to claim 1, wherein described tentatively identify alarm thing according to the warning information
Part forms preliminary identification data, comprising:
The warning information is tentatively identified into alarm event by rule base, forms preliminary identification data.
4. the method according to claim 1, wherein described to the warning information and the preliminary identification data
Pretreatment forms training sample set, comprising:
The warning information and the preliminary identification data are segmented and stop words is gone to handle；
Will treated warning information vectorization, form training sample set；Wherein, the announcement of any one event of the training sample set
Alert information is data matrix, the most major term length of the line number of the data matrix is that treated warning information, the data square
The columns of battle array is the dimension of treated warning information vectorization.
5. according to the method described in claim 4, it is characterized in that, it is described will treated warning information vectorization, form instruction
Practice sample set, comprising:
By treated, warning information passes through word2vec formation term vector, and the term vector constitutes the training sample set.
6. the method according to claim 1, wherein described use the training sample set to convolutional neural networks
Model training forms power system monitor alarm event identification model, comprising:
The warning information of input is normalized using the input layer of the convolutional neural networks model；
Normalized warning information is rolled up using multiple convolution units in the convolutional layer of the convolutional neural networks model
Product forms convolution characteristic pattern；
The convolution characteristic pattern maximum pondization is operated using the pond layer of the convolutional neural networks model, forms maximum value number
According to matrix；
The full articulamentum of the convolutional neural networks model is used to calculate the warning information of the input as the event of each definition
Probability, and choose the recognition result of the event of maximum probability as the warning information of the input, complete to the convolution mind
Through network model training, power system monitor alarm event identification model is formed.
7. a kind of device of power system monitor alarm event identification characterized by comprising
The warning information of power system monitor obtains module, for obtaining the warning information of power system monitor；
Preliminary identification data form module, and for tentatively identifying monitoring alarm event according to the warning information, formation is preliminary to be known
Other data；
Training sample set forms module, for forming training sample to the warning information and the preliminary identification data prediction
This collection；The training sample set includes the event of multiple definition；
Power system monitor alarm event identification model forms module, for using the training sample set to convolutional neural networks model
Training forms power system monitor alarm event identification model；
Power system monitor alarm event identification module, for identifying monitoring alarm using the power system monitor alarm event identification model
Event.
8. the device of power system monitor alarm event identification according to claim 7, which is characterized in that the preliminary identification number
It include regular library unit according to module is formed；
The rule library unit is used to the warning information tentatively identifying monitoring alarm event by rule base, is formed and tentatively known
Other data.
9. a kind of terminal device characterized by comprising
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Now the power system monitor alarm event as described in claim 1-6 is any knows method for distinguishing.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Realize that the power system monitor alarm event as described in claim 1-6 is any knows method for distinguishing when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910563871.7A CN110334756A (en) | 2019-06-26 | 2019-06-26 | Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910563871.7A CN110334756A (en) | 2019-06-26 | 2019-06-26 | Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110334756A true CN110334756A (en) | 2019-10-15 |
Family
ID=68142980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910563871.7A Pending CN110334756A (en) | 2019-06-26 | 2019-06-26 | Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334756A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111090747A (en) * | 2019-12-03 | 2020-05-01 | 国家电网有限公司 | Power communication fault emergency disposal method based on neural network classification |
| CN111404274A (en) * | 2020-04-29 | 2020-07-10 | 平顶山天安煤业股份有限公司 | Online monitoring and early warning system for displacement of power transmission system |
| CN111899122A (en) * | 2020-07-03 | 2020-11-06 | 国网江苏省电力有限公司镇江供电分公司 | User decentralized clearing method based on energy storage control |
| CN112446341A (en) * | 2020-12-07 | 2021-03-05 | 北京明略软件系统有限公司 | Alarm event identification method, system, electronic equipment and storage medium |
| CN113497729A (en) * | 2020-04-03 | 2021-10-12 | 中移动信息技术有限公司 | Training method of alarm recognition model, alarm recognition method and device |
| CN113743512A (en) * | 2021-09-07 | 2021-12-03 | 上海观安信息技术股份有限公司 | Autonomous learning judgment method and system for safety alarm event |
| CN113822421A (en) * | 2021-10-14 | 2021-12-21 | 平安科技（深圳）有限公司 | Neural network based anomaly positioning method, system, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108664374A (en) * | 2018-05-17 | 2018-10-16 | 腾讯科技（深圳）有限公司 | Fault warning model creation method, apparatus, fault alarming method and device |
| CN109117941A (en) * | 2018-07-16 | 2019-01-01 | 北京思特奇信息技术股份有限公司 | Alarm prediction method, system, storage medium and computer equipment |
-
2019
- 2019-06-26 CN CN201910563871.7A patent/CN110334756A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108664374A (en) * | 2018-05-17 | 2018-10-16 | 腾讯科技（深圳）有限公司 | Fault warning model creation method, apparatus, fault alarming method and device |
| CN109117941A (en) * | 2018-07-16 | 2019-01-01 | 北京思特奇信息技术股份有限公司 | Alarm prediction method, system, storage medium and computer equipment |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111090747A (en) * | 2019-12-03 | 2020-05-01 | 国家电网有限公司 | Power communication fault emergency disposal method based on neural network classification |
| CN113497729A (en) * | 2020-04-03 | 2021-10-12 | 中移动信息技术有限公司 | Training method of alarm recognition model, alarm recognition method and device |
| CN111404274A (en) * | 2020-04-29 | 2020-07-10 | 平顶山天安煤业股份有限公司 | Online monitoring and early warning system for displacement of power transmission system |
| CN111899122A (en) * | 2020-07-03 | 2020-11-06 | 国网江苏省电力有限公司镇江供电分公司 | User decentralized clearing method based on energy storage control |
| CN111899122B (en) * | 2020-07-03 | 2024-01-02 | 国网江苏省电力有限公司镇江供电分公司 | User decentralized clearing method based on energy storage control |
| CN112446341A (en) * | 2020-12-07 | 2021-03-05 | 北京明略软件系统有限公司 | Alarm event identification method, system, electronic equipment and storage medium |
| CN113743512A (en) * | 2021-09-07 | 2021-12-03 | 上海观安信息技术股份有限公司 | Autonomous learning judgment method and system for safety alarm event |
| CN113822421A (en) * | 2021-10-14 | 2021-12-21 | 平安科技（深圳）有限公司 | Neural network based anomaly positioning method, system, equipment and storage medium |
| CN113822421B (en) * | 2021-10-14 | 2024-05-14 | 平安科技（深圳）有限公司 | Neural network-based anomaly locating method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334756A (en) | Power system monitor alarm event knows method for distinguishing, terminal installation, equipment and medium | |
| Zhang et al. | Analog circuit incipient fault diagnosis method using DBN based features extraction | |
| CN103914735B (en) | A kind of fault recognition method and system based on Neural Network Self-learning | |
| CN111832608B (en) | Iron spectrum image multi-abrasive particle identification method based on single-stage detection model yolov3 | |
| Khormali et al. | A novel approach for recognition of control chart patterns: Type-2 fuzzy clustering optimized support vector machine | |
| CN112838946B (en) | Method for constructing intelligent sensing and early warning model based on communication network faults | |
| CN113255848A (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| CN106874963B (en) | A kind of Fault Diagnosis Method for Distribution Networks and system based on big data technology | |
| CN112434390B (en) | PCA-LSTM bearing residual life prediction method based on multi-layer grid search | |
| CN106482967A (en) | A kind of Cost Sensitive Support Vector Machines locomotive wheel detecting system and method | |
| CN116738339A (en) | Multi-classification deep learning recognition detection method for small-sample electric signals | |
| CN110737976A (en) | mechanical equipment health assessment method based on multi-dimensional information fusion | |
| CN110851654A (en) | Industrial equipment fault detection and classification method based on tensor data dimension reduction | |
| Remadna et al. | An overview on the deep learning based prognostic | |
| Guang et al. | Multi-scale RCNN model for financial time-series classification | |
| Setiono et al. | Using sample selection to improve accuracy and simplicity of rules extracted from neural networks for credit scoring applications | |
| CN113221442B (en) | Method and device for constructing health assessment model of power plant equipment | |
| CN112163474B (en) | Intelligent gearbox diagnosis method based on model fusion | |
| CN116860529A (en) | Fault positioning method and device | |
| CN113884807A (en) | Power distribution network fault prediction method based on random forest and multi-layer architecture clustering | |
| CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
| CN116452335A (en) | Financial big data management system based on time sequence neural network | |
| Ge et al. | Remaining useful life prediction using deep multi-scale convolution neural networks | |
| CN115713027A (en) | Transformer state evaluation method, device and system | |
| CN114861977A (en) | Distillation integration mode perception algorithm and system for unbalanced power data |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191015 |