CN114841199A - Power distribution network fault diagnosis method, device, equipment and readable storage medium - Google Patents
Power distribution network fault diagnosis method, device, equipment and readable storage medium Download PDFInfo
- Publication number
- CN114841199A CN114841199A CN202210406063.1A CN202210406063A CN114841199A CN 114841199 A CN114841199 A CN 114841199A CN 202210406063 A CN202210406063 A CN 202210406063A CN 114841199 A CN114841199 A CN 114841199A
- Authority
- CN
- China
- Prior art keywords
- power distribution
- network
- distribution network
- diagnosis
- fault diagnosis
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 137
- 238000009826 distribution Methods 0.000 title claims abstract description 121
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000011176 pooling Methods 0.000 claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 46
- 230000008569 process Effects 0.000 claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 21
- 230000004913 activation Effects 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 56
- 238000010606 normalization Methods 0.000 claims description 17
- 210000002569 neuron Anatomy 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 abstract description 11
- 238000010586 diagram Methods 0.000 description 20
- 238000013527 convolutional neural network Methods 0.000 description 17
- 238000012360 testing method Methods 0.000 description 9
- 239000011159 matrix material Substances 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 7
- 238000004088 simulation Methods 0.000 description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 101100499229 Mus musculus Dhrsx gene Proteins 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- 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
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Signal Processing (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention provides a power distribution network fault diagnosis method, a power distribution network fault diagnosis device, power distribution network fault diagnosis equipment and a readable storage medium. The method comprises the following steps: constructing a basic diagnosis network, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer; adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in a basic diagnosis network to obtain a to-be-trained diagnosis network; preprocessing fault current information collected by a fault recorder to obtain a sample data set; training a to-be-trained diagnosis network through a sample data set to obtain a power distribution network fault diagnosis network, and performing supervision training by taking FocalLoss as a target function in the training process; and carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network. By the invention, the calculation amount is reduced and the diagnosis accuracy is improved.
Description
Technical Field
The invention relates to the technical field of machine learning, in particular to a power distribution network fault diagnosis method, device, equipment and a readable storage medium.
Background
As a final phase of the power transmission, the distribution network is composed of a plurality of branches and delivers the electric energy to the final consumption point. The branches are scattered in vast rural areas and cities and are susceptible to different types of faults caused by different sources, such as severe weather conditions, vegetation growth, equipment faults and the like, and the data show that 80% of the customer power utilization interruptions are caused by power distribution network faults. The duration of the power interruption is one of the most important reliability indicators for the utility company, and it is therefore very important to diagnose faults quickly and efficiently to minimize downtime. The nonuniformity of the power distribution network, the unbalance of the structure and the circuit branch prevent the direct application of the fault diagnosis method of the power transmission network on the power distribution network.
In recent years, different fault diagnosis techniques have been explored and proposed to address these challenges, mainly expert systems, artificial neural networks, analytical models, etc. The expert system utilizes experience knowledge of experts and related theoretical knowledge to establish a rule base to carry out logic matching with the power grid fault so as to diagnose the fault, accords with the thinking and cognitive process of human beings, and has strong reasoning and interpretation capability. But at the same time, the expert system needs to update the rule base in real time, and the maintenance is difficult. The artificial neural network can automatically learn and adjust network parameters through sample training, input fault information and output diagnosed fault types. However, this method requires a large amount of sample data to train the network, it is difficult to actually obtain a large amount of complete fault sample data, and when the grid structure changes, the network needs to be trained and learned again.
The Petri network diagnoses the faults of the power distribution network by adopting weighted directed graph and matrix operation, and deduces and analyzes the logical relation among power grid elements, protection and circuit breakers by using a mathematical method. However, the existing Petri net model needs to individually model all the fault elements by adopting a mode of off-line modeling and on-line calling, which is not suitable for all networks. The analytical model is a mathematical model established based on a protection fault hypothesis and expected states of circuit breakers in the system, and converts a fault diagnosis problem into an integer programming solution. However, the traditional analytic model is easy to cause multiple solutions and misunderstanding, in a large complex system, the mechanism of the system is difficult to analyze and model, and the variable dimension is increased, so that the solution is increasingly difficult.
Most of the existing mainstream power distribution network diagnosis methods perform feature extraction manually, have certain subjectivity, can only express the specified features of fault signals, and are inconvenient to transplant to other data sets for direct use due to the nonuniformity, unbalanced structure and circuit branches of the power distribution network.
Disclosure of Invention
In order to solve the technical problem, the invention provides a power distribution network fault diagnosis method, a device, equipment and a readable storage medium.
In a first aspect, the present invention provides a power distribution network fault diagnosis method, where the power distribution network fault diagnosis method includes:
constructing a basic diagnosis network, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer;
adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in a basic diagnosis network to obtain a to-be-trained diagnosis network;
preprocessing fault current information collected by a fault recorder to obtain a sample data set;
training a to-be-trained diagnosis network through a sample data set to obtain a power distribution network fault diagnosis network, and performing supervision training by taking Focal local as a target function in the training process;
and carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
Optionally, the step of preprocessing the fault current information collected by the fault recorder to obtain a sample data set includes:
importing fault current information collected by a fault recorder into MATLAB software, and obtaining a sample data set by using a max function and a for loop, wherein the fault current information is represented as follows:
x＝[Time I a I b I c FT FR FIA LV FL]
wherein, I a Denotes the phase of A current, I b Denotes phase B current, I c Representing C phase current, FT representing fault type, FR representing fault resistance, FIA representing fault initial phase angle, LV representing load size, FL representing fault position;
the sample data set is represented as follows:
where 1 to n are the labels of the fault current information 1 to n.
Optionally, the residual structure module is simplified based on the number of convolutional layers.
Optionally, the batch normalization layer is configured to normalize the input value distribution of any neuron to a standard normal distribution.
Optionally, the shuffle net module is shuffle net _ V2, and the shuffle net _ V2 divides the input feature into two branches in the channel dimension, where one branch remains unchanged and performs the same mapping, and the other branch performs three times of convolution with the same output and input channel.
Optionally, the attention module comprises a channel attention module, and the channel attention module performs a maximum pooling operation first to obtain a pooling result M c Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A c Assuming that the number of channels of the input feature is represented by c, M is c And A c Respectively input the full connection layer and output M c ' and A c ' after summing, inputting a sigmoid function to carry out standard normalization, and outputting a result of the sigmoid function as channel attention.
Optionally, the attention module further includes a spatial attention module, and the spatial attention module performs a maximum pooling operation on each channel to obtain a pooling result M s Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A s The length of the input feature is represented by sSplicing pooling results [ M s ,A s ]To [ M ] s ,A s ]Performing a convolution kernel of size [1,7,2 ]]And performing convolution operation with the number of convolution kernels being 1, inputting a convolution result into a sigmoid function for standard normalization, and outputting a result of the sigmoid function as space attention.
In a second aspect, the present invention further provides a power distribution network fault diagnosis apparatus, including:
the building module is used for building a basic diagnosis network, and the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer;
the optimization module is used for adding a residual error structure module, a batch normalization layer, a ShuffleNet module and an attention module in the basic diagnosis network to obtain a to-be-trained diagnosis network;
the data set construction module is used for preprocessing fault current information collected by the fault recorder to obtain a sample data set;
the training module is used for training the diagnostic network to be trained through the sample data set to obtain the fault diagnostic network of the power distribution network, and the Focal local is taken as a target function to perform supervision training in the training process;
and the diagnosis module is used for carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
In a third aspect, the present invention further provides a power distribution network fault diagnosis apparatus, which includes a processor, a memory, and a power distribution network fault diagnosis program stored on the memory and executable by the processor, wherein the power distribution network fault diagnosis program, when executed by the processor, implements the steps of the power distribution network fault diagnosis method described above.
In a fourth aspect, the present invention further provides a readable storage medium, where the readable storage medium stores a power distribution network fault diagnosis program, where the power distribution network fault diagnosis program, when executed by a processor, implements the steps of the power distribution network fault diagnosis method described above.
In the invention, a basic diagnosis network is constructed, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer; adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in a basic diagnosis network to obtain a to-be-trained diagnosis network; preprocessing fault current information collected by a fault recorder to obtain a sample data set; training a to-be-trained diagnosis network through a sample data set to obtain a power distribution network fault diagnosis network, and performing supervision training by taking Focal local as a target function in the training process; and carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network. According to the method, the original power grid current data (namely fault current information collected by a fault recorder) is used as input, a fault diagnosis result is output, manual participation is not needed, the method has the characteristics of high detection precision and good data set portability, a simplified residual error connection technology is used, Focal Loss is used as a target function for supervision training, a batch standardization layer is added in a network for optimization, and a flow based on ShuffleNet network reduction parameters and attention mechanism improvement precision is adopted, so that a proper CNN framework can be automatically determined aiming at the fault diagnosis problem. The CNN adopted in the invention directly takes the current information obtained by up-sampling of the fault recorder of each node in the power distribution network as input, does not need to extract frequency domain characteristics and manual selection characteristics by using a digital signal processing method, and effectively improves the diagnosis precision by combining the advantages of the ShuffleNet module and the attention module.
Drawings
Fig. 1 is a schematic hardware structure diagram of a power distribution network fault diagnosis device according to an embodiment of the present invention;
fig. 2 is a schematic flow chart of an embodiment of a power distribution network fault diagnosis method according to the present invention;
fig. 3 is a residual error structure diagram adopted in an embodiment of the power distribution network fault diagnosis method of the present invention;
fig. 4 is a simplified residual error structure diagram adopted in an embodiment of the power distribution network fault diagnosis method of the present invention;
FIG. 5 is a schematic diagram of the addition of residual structure blocks in the basic diagnostic network;
FIG. 6 is a schematic diagram of the addition of a ShuffleNet module to a basic diagnostic network;
FIG. 7 is a process flow diagram of a channel attention module;
FIG. 8 is a process flow diagram of a spatial attention module;
FIG. 9 is a process flow diagram of an attention module;
FIG. 10 is a schematic diagram of the addition of a ShuffleNet module and an attention module to a basic diagnostic network;
fig. 11 is a functional module schematic diagram of an embodiment of a power distribution network fault diagnosis apparatus according to the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In a first aspect, an embodiment of the present invention provides a power distribution network fault diagnosis device, which may be a device with a data processing function, such as a Personal Computer (PC), a laptop, a server, and the like.
Referring to fig. 1, fig. 1 is a schematic diagram of a hardware structure of a power distribution network fault diagnosis device according to an embodiment of the present invention. In this embodiment of the present invention, the power distribution network fault diagnosis device may include a processor 1001 (e.g., a Central Processing Unit, CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used for realizing connection communication among the components; the user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard); the network interface 1004 may optionally include a standard wired interface, a WIreless interface (e.g., a WI-FI interface, WI-FI interface); the memory 1005 may be a Random Access Memory (RAM) or a non-volatile memory (non-volatile memory), such as a magnetic disk memory, and the memory 1005 may optionally be a storage device independent of the processor 1001. Those skilled in the art will appreciate that the hardware configuration depicted in FIG. 1 is not intended to be limiting of the present invention, and may include more or less components than those shown, or some components in combination, or a different arrangement of components.
With continued reference to fig. 1, a memory 1005, which is one type of computer storage medium in fig. 1, may include an operating system, a network communication module, a user interface module, and a power distribution network fault diagnostic program. The processor 1001 may call the power distribution network fault diagnosis program stored in the memory 1005, and execute the power distribution network fault diagnosis method provided by the embodiment of the present invention.
In a second aspect, an embodiment of the present invention provides a power distribution network fault diagnosis method.
In an embodiment, referring to fig. 2, fig. 2 is a schematic flowchart of an embodiment of a power distribution network fault diagnosis method according to the present invention. As shown in fig. 2, the power distribution network fault diagnosis method includes:
step S10, constructing a basic diagnosis network, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer;
in this embodiment, constructing the basic diagnostic network includes:
constructing a rolling layer:
the convolutional layer is composed of a set of convolutional kernels for extracting different features, and the mathematical expression of the calculation is as follows:
wherein the input characteristic diagram of the convolutional layer is represented by M j It is shown that the bias value is represented by b, the nonlinear activation function is represented by f (-), and the input feature map of the i-th convolution kernel of the l-1-th layer network is represented by
The jth weight value of ith convolution kernel of the ith layer is expressed by
Is expressed by the symbol of convolution operation, and finally, the jth characteristic diagram of the ith layer is output, namely
Constructing a pooling layer:
there are two common pooling operations: average Pooling operation (Average Pooling) and maximum Pooling operation (Max Pooling). The general computational process of pooling can be expressed as:
in the formula, the multiplicative bias of the jth neuron of the l-th network is formed by
Indicating that the additive bias of the jth neuron of the l-th network is
The sampling function is denoted by down (-).
Constructing a full connection layer:
the fully-connected layer input is a feature extracted by the layer in front of the network, and the output is a one-dimensional vector. The mathematical expression is as follows:
y＝W d ·x+b d
in which x denotes the input of the full link layer, W d Is a parameter matrix, b d Is a bias vector.
Constructing a Relu activation function:
the activation function introduces non-linearities into the computation of the neural network. The nonlinear activation functions of the output layer and the hidden layer adopted in the embodiment are Relu functions.
Constructing a neural network hyper-parameter;
through a large number of tests, the Adam algorithm is more suitable for fault diagnosis of the convolutional neural network power distribution network. The Adam algorithm can be expressed as follows using a mathematical formula:
μ dα ＝α 1 μ dα +(1-γ 1 )dα
Y dα ＝γ 2 Y dα +(1-γ 2 )(dα) 2
in the formula, the deviation of the exponential weighted average is corrected by the first two formulas, and the hyperparameter of the Adam algorithm is composed of mu and gamma 1 、γ 2 Is represented by gamma 1 Typically 0.9, and gamma 2 Typically 0.999.
Through a large number of tests, the Adam algorithm is more suitable for fault diagnosis of the convolutional neural network power distribution network, and the used target function is a cross entropy loss function.
Setting a training sample set as alpha { (alpha) 1 ,β 1 ),(α 2 ,β 2 ),...,(α n ,β n ) B, β ∈ {0,1}, then the mathematical expression of the cross-entropy loss function is:
in the formula, the probability that the sample is positive is represented by beta E [0,1 ∈]Is represented by the formula p t I.e. can be represented as:
it is possible to obtain:
CE(p,β)＝CE(p t )＝-log(p t )
if N is used to represent the batch size in the training process, the loss in one iteration can be represented as:
in the formula, the probability that the ith sample belongs to the positive class is determined by
And (4) showing.
Constructing a Softmax layer:
taking the ith node output as an example, the mathematical definition of the Softmax function is:
in the formula, z i And C is the output value of the ith node, and the number of output nodes, namely the number of classified categories. The output value of the multi-classification can be converted into the range of [0,1 ] through the Softmax function]The probability distribution of (c).
And finally, combining the input layer, the convolution layer, the pooling layer, the full-link layer, the activation function, the neural network hyper-parameter and the softmax layer to obtain the basic diagnostic network.
Step S20, adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in the basic diagnosis network to obtain a diagnosis network to be trained;
in the embodiment, in order to solve the problems of network degradation and gradient disappearance caused by the fact that the depth of the convolutional neural network is too deep, a residual error structure module is added in the basic diagnosis network. This embodiment may employ a residual structure implemented by "jump chaining", as shown in fig. 3.
Further, in one embodiment, the residual structure module is simplified based on the number of convolutional layers.
In this embodiment, considering that only four convolutional layers are needed for power distribution network fault diagnosis, the embodiment can also simplify the structure of the classical residual error structure unit, as shown in fig. 4, the convolutional layers are used for simple feature extraction, and then "layer jump connection" is used for constructing identity mapping.
Referring to fig. 5, fig. 5 is a schematic diagram of adding a residual structure module in the basic diagnostic network.
In this embodiment, a batch normalization layer is further added to the basic diagnostic network.
Further, in one embodiment, the batch normalization layer is configured to normalize the input value distribution of any neuron to a standard normal distribution.
In this embodiment, adding the batch normalization layer normalizes the input value distribution of any neuron into a standard normal distribution, and in each layer of the neural network model, the average value is 0 and the variance is 1.
In order to reduce the calculation amount, a ShuffleNet module is also added in the basic diagnosis network.
Further, in an embodiment, the shuffle module is shuffle _ V2, shuffle _ V2 divides the input feature into two branches in the channel dimension, wherein one branch remains unchanged and performs the same mapping, and the other branch performs three consecutive convolutions with the same output and input channel.
In this embodiment, ShuffleNet _ V2 is used. Shuffle _ V2 splits the input feature into two branches in the channel dimension, the channel numbers of the two branches being c-c ' and c ', respectively, where c ' is c/2. One of the branches remains unchanged and performs the same mapping, while the other branch performs three successive convolutions with the same output and input channels;
the ShuffleNet module can greatly reduce the calculation amount by reducing the parameter amount of the network model. Tests prove that the calculated amount is reduced to 36.104% after the ShuffleNet module is introduced, and the accuracy rate is only reduced by 0.00073%. Referring to fig. 6, fig. 6 is a schematic diagram of adding a shuffle net module to the basic diagnostic network.
In order to further improve the diagnosis accuracy, an attention module is also added in the basic diagnosis network.
Further, in one embodiment, the attention module includes a channel attention module, and the channel attention module first performs a maximum pooling operation to obtain a pooling result M c Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A c Assuming that the number of channels of the input feature is represented by c, M is c And A c Respectively input the full connection layer and output M c ' and A c ', after summing, inputting a sigmoid function for standard normalization, and outputting a result of the sigmoid function as channel attention. Referring to fig. 7, fig. 7 is a process flow diagram of the channel attention module.
Further, in an embodiment, the attention module further includes a spatial attention module, and the spatial attention module first performs a maximal pooling operation on each channel to obtain a pooling result M s Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A s The length of the input features is represented by s, and the pooling result is spliced [ M s ,A s ]To [ M ] s ,A s ]Performing a convolution kernel of size [1,7,2 ]]And performing convolution operation with the number of convolution kernels being 1, inputting a convolution result into a sigmoid function for standard normalization, and outputting a result of the sigmoid function as space attention. Referring to fig. 8, fig. 8 is a process flow diagram of the spatial attention module.
Assume that the input to the network is X N ＝[x 1 ,x 2 ,…,x N ]Representing the original network by f H (. -) shows the joined feedforward network is represented by A Expressed, then the output of the feed-forward network added by the original network model and attention mechanism can be expressed as:
H N ＝(h 1 ,h 2 ,…,h N )＝f N (X N )
A N ＝(a 1 ,a 2 ,…,a N )＝f A (X N ,Q)
in the equation, the other input calculation of the attention mechanism is represented by Q. H is to be N And A N The combination is the final output of the attention module, usually by means of a weighted sumBinding H N And A N The final output Y of the attention module can then be expressed as:
the attention module (CBAM) passes the input features through the channel attention module and then through the spatial attention module. The simplified residual error connection technology constructed in the embodiment can be obtained by introducing the attention module to obtain the optimized convolutional neural network, supervised training is carried out by taking Focal local as a target function, a BatchNormalization layer is added in the network for optimization, and the precision of the convolutional neural network (namely, the diagnostic network to be trained) is improved based on ShuffleNet network reduction parameters and the attention mechanism. Through tests, the diagnosis accuracy can be improved to more than 99% after the ShuffleNet module is introduced. Referring to fig. 9, fig. 9 is a process flow diagram of the attention module.
Referring to fig. 10, fig. 10 is a schematic diagram of adding a shuffle net module and an attention module to the basic diagnostic network.
Step S30, preprocessing the fault current information collected by the fault recorder to obtain a sample data set;
in this embodiment, the fault current information collected by the fault recorder is preprocessed to obtain a sample data set for a subsequent training process.
Further, in one embodiment, step S30 includes:
importing fault current information collected by a fault recorder into MATLAB software, and obtaining a sample data set by using a max function and a for loop, wherein the fault current information is represented as follows:
x＝[Time I a I b I c FT FR FIA LV FL]
wherein, I a Denotes the phase of A current, I b Denotes the phase B current, I c Representing C phase current, FT representing fault type, FR representing fault resistance, FIA representing fault initial phase angle, LV representing load size, FL representing fault position;
the sample data set is represented as follows:
where 1 to n are the labels of the fault current information 1 to n.
Step S40, training the diagnostic network to be trained through the sample data set to obtain a power distribution network fault diagnostic network, and performing supervision training by taking Focal local as a target function in the training process;
in the embodiment, the sample data set can be divided into a training set, a verification set and a test set according to a proportion, so that the to-be-trained diagnostic network is trained to obtain the power distribution network fault diagnosis network, and in the training process, the Focal local is taken as a target function for supervision training.
The mathematical expression of Focal local is as follows:
FL＝-x t (1-p t ) y log(p t )
in the formula, the hyperparameters are represented by x and y, and the prediction probability of the sample is represented by p t Expressed by a modulation factor of (1-p) t ) y And (4) showing. When y is more than or equal to 1, the more accurate a sample is predicted, the more accurate p it is t The closer to 1, the modulation factor (1-p) t ) y The smaller, the less accurate a sample is predicted, its p t The closer to 0, the modulation factor (1-p) t ) y The larger.
And step S50, carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
In this embodiment, it should be noted that, when performing fault diagnosis, data input to the power distribution network fault diagnosis network is fault current information collected by the fault recorder.
In the embodiment, a basic diagnosis network is constructed, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer; adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in a basic diagnosis network to obtain a to-be-trained diagnosis network; preprocessing fault current information collected by a fault recorder to obtain a sample data set; training a to-be-trained diagnosis network through a sample data set to obtain a power distribution network fault diagnosis network, and performing supervision training by taking Focal local as a target function in the training process; and carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network. According to the method, original power grid current data (namely fault current information collected by a fault recorder) is used as input, a fault diagnosis result is output, manual participation is not needed, the method has the characteristics of high detection precision and good data set portability, a simplified residual error connection technology is used, Focal Loss is used as a target function for supervision training, a batch standardization layer is added in a network for optimization, and a proper CNN architecture can be automatically determined based on processes of reducing parameters of a ShuffleNet network and improving precision based on an attention mechanism. The CNN adopted in the embodiment directly takes current information obtained by up-sampling of a fault recorder of each node in the power distribution network as input, does not need to extract frequency domain characteristics and manual selection characteristics by using a digital signal processing method, and effectively improves the diagnosis precision by combining the advantages of a ShuffleNet module and an attention module.
Further, in one embodiment, the experiment is carried out by using a simulation model of the IEEE34 power distribution network built in PSCAD. And a detailed comparison of the performance with other existing methods. In order to restore the operation condition of the actual power distribution network as much as possible, so that the samples for network training are rich enough, possible fault factors need to be considered comprehensively, and a fault recorder needs to be simulated near each node.
The power distribution network system runs at the frequency of 24.9kV and 50Hz, and the sampling frequency is 2000 Hz. The Multiple Run components in PSCAD were used to generate training data for different fault types, fault resistances, initial phase angles, load sizes, and fault locations on transmission lines 808-.
TABLE 1 simulation data set configuration for fault diagnosis of power distribution networks
A total of 27648 three-phase current raw simulation datasets containing fault signature parameter tags were generated and simulated for 4 fault lines. And importing the original simulation data set into MATLAB, obtaining the maximum value of all three-phase currents and a fault characteristic parameter label matrix by using a max function and for circulation, and arranging the maximum value and the fault characteristic parameter label matrix into a one-dimensional vector as shown in the following as the input of a neural network.
To perform cross-validation and prevent the overfitting problem, all simulation data were compared as 6: 2: the scale of 2 is divided into a training set, a validation set, and a test set. The training set is used for training parameters of the convolutional neural network and adjusting a network structure, and the verification set is used for avoiding the overfitting problem in the training and network self-adaptive adjustment process. The test set is used to evaluate the performance of the network fabric. The raw fault current data set is shown in table 2 and the pre-processed current data set is shown in table 3.
TABLE 2 original Fault Current data set
TABLE 3 Pre-processed Fault Current dataset
The super-parameters of the network are debugged by adopting a Grid Search method, and the obtained super-parameters suitable for the convolutional neural network introduced with the ShuffleNet module and the convolutional attention module are shown in the table 4.
Table 4 Power distribution network fault diagnosis parameters introducing attention mechanism and ShuffleNet module
The Artificial Neural Network (ANN), the non-optimized Convolutional Neural Network (CNN), the optimized convolutional neural network (opt-CNN), the optimized neural network (opt-ShuCNN) introduced with the Shufflent module, the optimized power distribution network fault diagnosis convolutional neural network (opt-attention-ShuCNN) introduced with the Shufflenet module and the attention module are used for power distribution network fault diagnosis, results obtained on a test set are comprehensively compared, the accuracy rates of the five models are shown in the table 5, the average accuracy is shown in the table 5, and the average accuracy is shown in the table
The average recall is shown in table 6, and the average F1 value is shown in table 7 and table 8, respectively.
TABLE 5 Total model accuracy
TABLE 6 average accuracy of all models
TABLE 7 average recall rate for all models
TABLE 8 ANN, CNN, opt-CNN model mean F1
Note: average accuracy, average recall, and average F1 values were obtained on the test set.
It can be seen that the accuracy of the artificial neural network is the lowest and only 93.926% after the artificial neural network is replaced by the convolutional neural network, the accuracy is improved greatly after a series of optimization is performed on the convolutional neural network, the accuracy is improved remarkably, but the parameter quantity of the network is increased obviously, the calculation complexity is increased, and then a ShuffleNet module is introduced, so that the parameter quantity is reduced greatly, the calculation quantity is reduced greatly, but the calculation precision is reduced slightly, and finally the convolutional attention module is introduced into the network.
The obtained channel attention is classified by the second convolution attention module according to the running state of the power distribution network, and the channel attention mean value of each power distribution network running state is visualized, so that the channel attention in the network model can reflect the importance of each channel to a certain extent, and the running state of the power distribution network can be expressed by the importance of each channel in the network model, which shows that the current characteristic of the power distribution network has strong relevance with the running state, and the power distribution network fault diagnosis by using the convolution neural network is feasible and the principle of the power distribution network fault diagnosis is interpretable.
Further, the evaluation indexes of the prediction result in this embodiment are:
1) confusion matrix
By comparing the prediction results of the samples with the true labels, the samples are classified in the classification problem, so that the number of the samples with correct classification and the number of the samples with wrong classification can be obtained. The confusion matrix of the classification result can be further obtained by arranging the number of the correctly classified and the wrong samples, as shown in table 9:
TABLE 9 confusion matrix for classification results
Wherein TP indicates that the sample is classified correctly and belongs to the original label category; TN indicates that the sample is classified correctly, but not in the original label category; FN indicates that the sample is classified as erroneous, but belongs to the original label category; FP indicates that the sample was misclassified and does not belong to the original label category.
2) Precision (Precision)
The accuracy refers to the proportion of the number of samples correctly judged as a certain fault category by the classifier to the total number of samples judged as the fault category by the classifier, and can reflect the accuracy of the classification of the constructed model, and the calculation formula is as follows:
3) recall ratio (Recall)
The recall rate is the proportion of the number of samples of a certain fault category correctly judged by the classifier in the total number of texts belonging to the category, and the calculation formula is as follows:
4) f1 value
The F1 value is the performance index after the balance of the comprehensive accuracy and the recall rate, and the calculation formula is as follows:
each of these three performance metrics has its own Micro (Micro), Macro (Macro), and Weight (Weight) metrics. Micro makes statistics on each sample non-categorical in the dataset to build a global confusion matrix, and then calculates corresponding indexes. Macro is used for counting whether the classification of the overall data is correct or not and calculating a corresponding evaluation index. Weight is used for counting the proportion of the number of samples of each category to the total number of samples of all categories as Weight, and calculating corresponding evaluation indexes.
In a third aspect, the embodiment of the invention further provides a power distribution network fault diagnosis device.
In an embodiment, referring to fig. 11, fig. 11 is a functional module schematic diagram of an embodiment of a power distribution network fault diagnosis device according to the present invention. As shown in fig. 11, the power distribution network fault diagnosis apparatus includes:
the building module 10 is used for building a basic diagnosis network, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer;
the optimization module 20 is configured to add a residual error structure module, a batch normalization layer, a ShuffleNet module, and an attention module to the basic diagnostic network to obtain a diagnostic network to be trained;
the data set construction module 30 is configured to preprocess the fault current information collected by the fault recorder to obtain a sample data set;
the training module 40 is used for training the network to be trained and diagnosed through the sample data set to obtain the power distribution network fault diagnosis network, and during the training process, the Focal local is used as a target function for supervision and training;
and the diagnosis module 50 is used for carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
Further, in an embodiment, the data set constructing module 30 is configured to:
importing fault current information collected by a fault recorder into MATLAB software, and obtaining a sample data set by using a max function and a for loop, wherein the fault current information is represented as follows:
x＝[Time I a I b I c FT FR FIA LV FL]
wherein, I a Denotes the phase of A current, I b Denotes the phase B current, I c Representing C phase current, FT representing fault type, FR representing fault resistance, FIA representing fault initial phase angle, LV representing load size, FL representing fault position;
the sample data set is represented as follows:
where 1 to n are the labels of the fault current information 1 to n.
Further, in one embodiment, the residual structure module is simplified based on the number of convolutional layers.
Further, in one embodiment, the batch normalization layer is configured to normalize the input value distribution of any neuron to a standard normal distribution.
Further, in an embodiment, the shuffle module is shuffle _ V2, shuffle _ V2 divides the input feature into two branches in the channel dimension, wherein one branch remains unchanged and performs the same mapping, and the other branch performs three consecutive convolutions with the same output and input channel.
Further, in one embodiment, the attention module includes a channel attention module, and the channel attention module first performs a maximum pooling operation to obtain a pooling result M c Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A c Assuming the number of channels of the input features is c Showing that Mc and Ac are respectively input into the full connection layer and output M c ' and A c ', after summing, inputting a sigmoid function for standard normalization, and outputting a result of the sigmoid function as channel attention.
Further, in an embodiment, the attention module further includes a spatial attention module, and the spatial attention module first performs a maximum pooling operation on each channel to obtain a pooling result M s Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A s The length of the input features is represented by s, and the pooling result is spliced [ M s ,A s ]To [ M ] s ,A s ]Performing a convolution kernel of size [1,7,2 ]]And performing convolution operation with the number of convolution kernels being 1, inputting a convolution result into a sigmoid function for standard normalization, and outputting a result of the sigmoid function as space attention.
The function implementation of each module in the power distribution network fault diagnosis device corresponds to each step in the power distribution network fault diagnosis method embodiment, and the functions and implementation processes are not described in detail here.
In a fourth aspect, the embodiment of the present invention further provides a readable storage medium.
The readable storage medium of the present invention stores a power distribution network fault diagnosis program, wherein the power distribution network fault diagnosis program, when executed by a processor, implements the steps of the power distribution network fault diagnosis method as described above.
The method implemented when the power distribution network fault diagnosis program is executed may refer to each embodiment of the power distribution network fault diagnosis method of the present invention, and details are not described here.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) as described above and includes instructions for causing a terminal device to execute the method according to the embodiments of the present invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. The power distribution network fault diagnosis method is characterized by comprising the following steps:
constructing a basic diagnosis network, wherein the basic diagnosis network comprises a convolution layer, a pooling layer, a full-connection layer, an activation function, a neural network hyper-parameter and a softmax layer;
adding a residual error structure module, a batch standardization layer, a ShuffleNet module and an attention module in a basic diagnosis network to obtain a to-be-trained diagnosis network;
preprocessing fault current information collected by a fault recorder to obtain a sample data set;
training a to-be-trained diagnosis network through a sample data set to obtain a power distribution network fault diagnosis network, and performing supervision training by taking Focal local as a target function in the training process;
and carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
2. The power distribution network fault diagnosis method according to claim 1, wherein the step of preprocessing the fault current information collected by the fault recorder to obtain a sample data set comprises:
importing fault current information collected by a fault recorder into MATLAB software, and obtaining a sample data set by using a max function and a for loop, wherein the fault current information is represented as follows:
x＝[Time I a I b I c FT FR FIA LV FL]
wherein, I a Denotes the phase of A current, I b Denotes the phase B current, I c Representing C phase current, FT representing fault type, FR representing fault resistance, FIA representing fault initial phase angle, LV representing load size, FL tableIndicating the fault position;
the sample data set is represented as follows:
where 1 to n are the labels of the fault current information 1 to n.
3. The power distribution network fault diagnosis method of claim 1, wherein the residual structure module is simplified based on the number of convolutional layers.
4. The method according to claim 1, wherein the batch normalization layer is configured to normalize the distribution of input values of any neuron to a standard normal distribution.
5. The power distribution network fault diagnosis method according to claim 1, characterized in that the shuffle net module is shuffle net _ V2, shuffle net _ V2 divides the input features into two branches in the channel dimension, wherein one branch remains unchanged and performs the same mapping, and the other branch performs three convolutions consecutively with the same output and input channels.
6. The method according to claim 1, wherein the attention module comprises a channel attention module, and the channel attention module performs a maximum pooling operation first to obtain a pooling result M c Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A c Assuming that the number of channels of the input feature is represented by c, M is c And A c Respectively input the full connection layer and output M c ' and A c ' after summing, inputting a sigmoid function to carry out standard normalization, and outputting a result of the sigmoid function as channel attention.
7. The method of fault diagnosis for distribution networks of claim 1, whereinThe attention module also comprises a space attention module which firstly carries out maximum pooling operation on each channel to obtain a pooling result M s Simultaneously carrying out average pooling operation on each channel to obtain a pooling result A s The length of the input features is represented by s, and the pooling result is spliced [ M s ,A s ]To [ M ] s ,A s ]Performing a convolution kernel of size [1,7,2 ]]And performing convolution operation with the number of convolution kernels being 1, inputting a convolution result into a sigmoid function for standard normalization, and outputting a result of the sigmoid function as space attention.
8. A power distribution network fault diagnosis device, characterized in that the power distribution network fault diagnosis device comprises:
the building module is used for building a basic diagnosis network, and the basic diagnosis network comprises a convolution layer, a pooling layer, a full-link layer, an activation function, a neural network hyper-parameter and a softmax layer;
the optimization module is used for adding a residual error structure module, a batch normalization layer, a ShuffleNet module and an attention module in the basic diagnosis network to obtain a to-be-trained diagnosis network;
the data set construction module is used for preprocessing fault current information collected by the fault recorder to obtain a sample data set;
the training module is used for training the network to be trained and diagnosed through the sample data set to obtain the power distribution network fault diagnosis network, and during training, supervised training is carried out by taking the Focal local as a target function;
and the diagnosis module is used for carrying out power distribution network fault diagnosis based on the power distribution network fault diagnosis network.
9. A power distribution network fault diagnosis device, characterized in that the power distribution network fault diagnosis device comprises a processor, a memory, and a power distribution network fault diagnosis program stored on the memory and executable by the processor, wherein the power distribution network fault diagnosis program, when executed by the processor, implements the steps of the power distribution network fault diagnosis method according to any one of claims 1 to 7.
10. A readable storage medium, characterized in that the readable storage medium stores thereon a power distribution network fault diagnosis program, wherein the power distribution network fault diagnosis program, when executed by a processor, implements the steps of the power distribution network fault diagnosis method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210406063.1A CN114841199A (en) | 2022-04-18 | 2022-04-18 | Power distribution network fault diagnosis method, device, equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210406063.1A CN114841199A (en) | 2022-04-18 | 2022-04-18 | Power distribution network fault diagnosis method, device, equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114841199A true CN114841199A (en) | 2022-08-02 |
Family
ID=82565004
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210406063.1A Pending CN114841199A (en) | 2022-04-18 | 2022-04-18 | Power distribution network fault diagnosis method, device, equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114841199A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115902615A (en) * | 2023-01-09 | 2023-04-04 | 佰聆数据股份有限公司 | Method and device for analyzing defects of power circuit breaker |
| CN116310599A (en) * | 2023-05-17 | 2023-06-23 | 湖北工业大学 | Power transformer fault diagnosis method and system based on improved CNN-PNN network |
| CN117011805A (en) * | 2023-10-07 | 2023-11-07 | 广东电网有限责任公司云浮供电局 | Data exception evaluation method, device, equipment and readable storage medium |
-
2022
- 2022-04-18 CN CN202210406063.1A patent/CN114841199A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115902615A (en) * | 2023-01-09 | 2023-04-04 | 佰聆数据股份有限公司 | Method and device for analyzing defects of power circuit breaker |
| CN116310599A (en) * | 2023-05-17 | 2023-06-23 | 湖北工业大学 | Power transformer fault diagnosis method and system based on improved CNN-PNN network |
| CN116310599B (en) * | 2023-05-17 | 2023-08-15 | 湖北工业大学 | Power transformer fault diagnosis method and system based on improved CNN-PNN network |
| CN117011805A (en) * | 2023-10-07 | 2023-11-07 | 广东电网有限责任公司云浮供电局 | Data exception evaluation method, device, equipment and readable storage medium |
| CN117011805B (en) * | 2023-10-07 | 2024-02-06 | 广东电网有限责任公司云浮供电局 | Data exception evaluation method, device, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wen et al. | A generalized remaining useful life prediction method for complex systems based on composite health indicator | |
| CN114841199A (en) | Power distribution network fault diagnosis method, device, equipment and readable storage medium | |
| Wang et al. | A compound framework for wind speed forecasting based on comprehensive feature selection, quantile regression incorporated into convolutional simplified long short-term memory network and residual error correction | |
| Guo et al. | Output-relevant variational autoencoder for just-in-time soft sensor modeling with missing data | |
| CN110119854A (en) | Voltage-stablizer water level prediction method based on cost-sensitive LSTM Recognition with Recurrent Neural Network | |
| CN101464964A (en) | Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis | |
| Wu et al. | A transformer-based approach for novel fault detection and fault classification/diagnosis in manufacturing: A rotary system application | |
| Yang et al. | Unsupervised feature learning for online voltage stability evaluation and monitoring based on variational autoencoder | |
| CN113177357B (en) | Transient stability assessment method for power system | |
| CN115689008A (en) | CNN-BilSTM short-term photovoltaic power prediction method and system based on ensemble empirical mode decomposition | |
| CN112116002A (en) | Determination method, verification method and device of detection model | |
| CN116245019A (en) | Load prediction method, system, device and storage medium based on Bagging sampling and improved random forest algorithm | |
| Chen et al. | Real‐time recognition of power quality disturbance‐based deep belief network using embedded parallel computing platform | |
| Ramirez-Gonzalez et al. | Convolutional neural nets with hyperparameter optimization and feature importance for power system static security assessment | |
| CN111079348A (en) | Method and device for detecting slowly-varying signal | |
| CN115017970A (en) | Migration learning-based gas consumption behavior anomaly detection method and system | |
| Cao et al. | Fast and explainable warm-start point learning for AC Optimal Power Flow using decision tree | |
| Huang et al. | Harnessing deep learning for population genetic inference | |
| CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
| Wang et al. | A transformer-based multi-entity load forecasting method for integrated energy systems | |
| CN110135592B (en) | Classification effect determining method and device, intelligent terminal and storage medium | |
| Dang et al. | seq2graph: Discovering dynamic non-linear dependencies from multivariate time series | |
| CN115965150A (en) | Power load prediction method based on weight distribution | |
| CN115936926A (en) | SMOTE-GBDT-based unbalanced electricity stealing data classification method and device, computer equipment and storage medium | |
| Kou et al. | Machine learning based models for fault detection in automatic meter reading systems |
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 |