CN116879672A - Intelligent identification method for total fault of area based on time convolution network - Google Patents
Intelligent identification method for total fault of area based on time convolution network Download PDFInfo
- Publication number
- CN116879672A CN116879672A CN202310636611.4A CN202310636611A CN116879672A CN 116879672 A CN116879672 A CN 116879672A CN 202310636611 A CN202310636611 A CN 202310636611A CN 116879672 A CN116879672 A CN 116879672A
- Authority
- CN
- China
- Prior art keywords
- fault
- network
- total
- convolution network
- identification
- 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 48
- 238000005259 measurement Methods 0.000 claims abstract description 39
- 239000011159 matrix material Substances 0.000 claims abstract description 27
- 238000004364 calculation method Methods 0.000 claims abstract description 15
- 238000010801 machine learning Methods 0.000 claims abstract description 10
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 238000013135 deep learning Methods 0.000 claims abstract description 8
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 230000000694 effects Effects 0.000 claims abstract description 7
- 230000007246 mechanism Effects 0.000 claims description 28
- 230000006870 function Effects 0.000 claims description 12
- 239000012634 fragment Substances 0.000 claims description 10
- 239000000284 extract Substances 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 5
- 230000010339 dilation Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000005406 washing Methods 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 6
- 238000012549 training Methods 0.000 description 13
- 230000002159 abnormal effect Effects 0.000 description 11
- 238000012360 testing method Methods 0.000 description 8
- 230000005856 abnormality Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 7
- 230000005611 electricity Effects 0.000 description 5
- 230000004927 fusion Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003064 k means clustering Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000004441 surface measurement Methods 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/085—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution lines, e.g. overhead
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A total fault intelligent identification method based on a time convolution network belongs to the field of electric energy metering. Extracting information in a measurement time sequence data matrix in a 2D convolution form; machine learning of transformer fault identification is performed by adopting a time convolution network; extracting and compressing high-dimensional features in an input data matrix to an output layer for further fault identification; adopting two different fault identification strategies, namely daily fault identification and annual fault identification; and (3) using a deep learning tool to directly locate the total table of the area with faults by carrying out automatic and batch analysis on the area monitoring data, and eliminating the interference of other unreasonable factors in the line or the area, thereby improving the economic operation rate of the line. The time series characteristics contained by each variable independently can be extracted, and meanwhile, the correlation among the variables can be transversely extracted; on the premise of ensuring the information extraction effect, the parameter scale of the network is greatly reduced, so that the calculation efficiency is improved.
Description
Technical Field
The invention belongs to the field of electric energy measurement, and particularly relates to an intelligent identification method for total fault of a platform area based on a time convolution network.
Background
The total table of the transformer area is related to the electricity quantity of the 10kV line electricity selling side and the electricity quantity of the transformer area electricity supplying side, so that the functions of starting up and down are achieved.
The electric energy abnormality caused by the total surface faults of the transformer area can simultaneously influence the calculation of 10kV and transformer area line loss, so that the reasonable rate of the line loss is reduced, the monitoring of the electric quantity of the transformer area is influenced, the load rate of the transformer area is misestimated, and the power supply reliability is not facilitated. The effective identification of the total fault of the area is the key for solving the problems.
Because the number of the line-associated transformer areas is large, the electric quantity of a single transformer area is small, fluctuation is not obvious, influence factors are complex and the like, the abnormal electric quantity of the transformer area is a common problem in 10kV line loss treatment, and whether the abnormal electric quantity is caused by the total fault of the transformer area is difficult to accurately judge.
At present, investigation is mainly performed by means of manual analysis and field verification, so that the accuracy is low and obvious labor cost is brought.
Common faults in the total table of the transformer area comprise abnormal collection, inconsistent field and file multiplying power, meter voltage and current losing, meter voltage and current polarity misconnection and the like.
At present, direct researches on total surface faults of a platform area are relatively few, and the direct researches mainly focus on analysis of line loss anomalies in the platform area.
For example, "GBDT-based electric power metering equipment fault prediction" (Liu Jinshuo, liu Bi, tense, et al, computer science, 2019,46 (S1): 392-396), a method for identifying a district metering equipment fault type and a method for predicting the life cycle thereof based on a gradient lifting decision tree (GBDT) are disclosed, and validity and advancement of a designed model are verified by adopting actual data, but the method only considers fault data, and the mining of current state operation data of metering equipment is insufficient; the intelligent ammeter fault prediction based on the multi-classification fusion model (Chen She, han Tong, wei Ling, etc.. Electric measurement and instrument (1-9)), provides a multi-classifier fusion method for identifying the fault of the ammeter in the transformer area, and a fusion algorithm is adopted to construct the multi-classifier after preprocessing fault data, wherein the fusion strategy is essentially a voting mechanism and does not consider the recognition capability of different classifiers on different faults; the association relation between each attribute and the fault type of the intelligent electric meter is analyzed in 'intelligent electric meter fault prediction technical research based on big data analysis' (Fan Shaohua. Beijing university of post, 2018.), an XGBoost platform area metering equipment fault prediction method based on weighted column sampling is provided, the effectiveness of the method is verified on electric meter fault data, but the data roughness selected by the method is higher, and the contained related attribute information is less.
In addition, "low-voltage transformer area line loss abnormality identification method based on k-means clustering algorithm" (Chen Hongtao, cai Hui, li Xiong, wang Ying, zheng Enhui, southern electric network technology, 2019,13 (02)) discloses that an unsupervised learning method such as k-means clustering is adopted to cluster original line loss data, so that fault reasons and line loss abnormality degrees corresponding to different categories are analyzed, and support is provided for overhaul decision; the method for identifying the line loss abnormality of the low-voltage area based on the deep neural network (Wang Haiyun, zhang Yan, furong, chen Yan, yang Liping, commonly known, zhang Zaichi, chen Qian, yuan Qingfang, power demand side management, 2018,20 (06): 31-35.) and the system for diagnosing the line loss abnormality of the low-voltage area based on the wireless communication and big data technology (Li Jianning, ma Xiaoli, yan Huamin, jiang Chen, electric power and energy source, 2019,40 (01): 36-40.) are disclosed, wherein the marked line loss data is classified or regressed by adopting a supervised learning algorithm such as the neural network, so as to realize the judgment of the line loss fault.
However, the degree of mining and utilization of the existing data by the above existing analysis method still needs to be improved, specifically:
1) The correlation exists among a plurality of measured variables such as voltage, current, power, three-phase unbalance rate and the like, the traditional identification method is often aimed at analyzing a single variable, and the correlation among the variables is not fully utilized to improve the fault identification precision;
2) Focusing and extracting key fragments in long sequence data. The data of the platform area is often kept relatively intact, and the length of the platform area can reach a plurality of years. In this rich history, however, the anomaly data may only exist in a few short segments. How to automatically extract effective fault related information from a large amount of normal data, analyze and utilize the effective fault related information, and consider the conciseness and efficiency of a model, and the effective fault related information is not solved well at present.
Disclosure of Invention
The invention aims to provide an intelligent identification method for total fault of a platform area based on a time convolution network. Aiming at the defect that the correlation between multiple measurement cannot be well extracted by the existing method, the time convolution network adopted by the technical scheme can efficiently extract the correlation of different measurement variables, so that the fault identification precision is improved. The method utilizes a deep learning tool, and can directly locate the total table of the failed station area by carrying out automatic and batch analysis on the station area Ts-run monitoring data, thereby eliminating the interference of other unreasonable factors in the line or the station area and further improving the economic operation rate of the line.
The technical scheme of the invention is as follows: the utility model provides a station area total surface fault intelligent identification method based on time convolution network, which comprises the collection and cleaning of the total surface measurement data of each transformer station area, and is characterized in that:
1) Establishing a transformer fault label;
2) Parallel measurement time sequence data of different variables are input into a network as a data matrix, and information in the data is extracted in a 2D convolution form;
3) Machine learning of transformer fault identification is performed by adopting a time convolution network;
4) Increasing the receptive field of the network by increasing the size of convolution kernels, increasing the number of stacks and increasing the expansion rate;
5) Extracting and compressing high-dimensional features in an input data matrix to an output layer for further fault identification;
6) On the basis of using a single attention function, a multi-head attention mechanism is adopted to promote the overall performance;
7) Adopting two different fault identification strategies, namely daily fault identification and annual fault identification;
8) Based on a time convolution network, high-dimensional features in a long sequence are efficiently extracted and fault identification is carried out;
9) Attention mechanism is introduced in fault identification of annual sequences to strengthen screening capability of key information fragments
10 By using a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of other unreasonable factors in the lines or the areas is eliminated, so that the economic operation rate of the lines is improved.
Specifically, the total surface measurement data of the transformer area is expressed as:
wherein E represents the number of measurement variables and L represents the length of the measurement sequence.
Further, the measured variables at least include voltage, current, and power.
Specifically, the transformer faults are identified as two classification problems, y=0 represents a normal state, and y=1 represents a fault state;
the machine learning of fault identification is expressed as y=f (X), that is, by using the machine learning model f and taking the measurement sequence as input, it is determined whether the transformer has a fault in the measurement time.
Specifically, the time convolution network is a full convolution network with an expansion structure;
let the convolution kernel be denoted as f= [ F 0 ,f 1 ,…,f M-1 ]Wherein M is the number of convolution kernels;
then for element X in the input matrix X m Is a dilation convolution operation of (a)Expressed as:
wherein d is the expansion ratio, and m-d.i is x m A pointer to a previous element;
when the expansion ratio d=1, the network is degraded into a conventional convolutional network;
when the expansion ratio d >1, the expansion convolution operation skips the element of (d-1)/d in the previous layer and focuses on the rest 1/d elements only;
the above characteristics of the expansion ratio can significantly improve the receptive field of the network and reduce the complexity of the network.
Further, the receptive field is calculated using the formula:
where K is the size of the convolution kernel, N stack Is the number of stacks of the network.
Specifically, according to the intelligent recognition method for the total fault of the platform area based on the time convolution network, on the basis of using a single attention function, multiple attention functions are adopted in parallel, so that the overall performance, namely a multi-head attention mechanism, can be improved;
the calculation of the multi-head attention mechanism is shown as follows:
MultiHead(Q,K,V)＝concate(head 1 ,head 2 ,...,head n )W o
head i ＝Attention(QW i Q ,KW i K ,VW i V )
wherein W is i Q ，W i K ，W i V And W is o Is a parameter matrix that can be learned during the mapping process.
Specifically, in the intelligent identification method of total fault in the area based on the time convolution network, a classical TCN structure is used in the identification of the daily degree abnormality; in annual fault identification, attention mechanisms are introduced to improve the extraction capacity for key information pieces, thereby improving the accuracy of the model.
According to the intelligent identification method for total fault of the area based on the time convolution network, measurement time sequence data of different variables are input into the network in parallel as a data matrix, and information in the data matrix is extracted in a 2D convolution mode; extracting time sequence features contained by each variable independently, and simultaneously transversely extracting correlations among the variables; on the basis, the unique expansion convolution structure of the time convolution network is utilized, and the parameter scale of the network is greatly reduced on the premise of guaranteeing the information extraction effect, so that the calculation efficiency is improved.
According to the intelligent identification method for the total fault of the area based on the time convolution network, the time convolution network is adopted to efficiently extract the correlation of different measured variables, so that the fault identification precision is improved; and by utilizing a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of lines or other unreasonable factors in the areas is eliminated, so that the economic operation rate of the lines is improved.
Compared with the prior art, the invention has the advantages that:
1. according to the technical scheme, two different fault identification strategies are adopted to verify the validity of the model, namely, daily fault identification and annual fault identification; the solar fault identification is used for judging whether the transformer is abnormal on the same day by reading a solar measurement value; the annual fault identification reads the measurement value of the whole year to judge whether the transformer has faults in the year;
2. according to the technical scheme, measurement time sequence data of different variables are input into a network in parallel as a data matrix, and information in the measurement time sequence data is extracted in a 2D convolution mode; the time sequence characteristics contained by each variable independently can be extracted, and meanwhile, the correlation among the variables can be transversely extracted; on the basis, the traditional deep convolution network is considered to have higher model complexity and calculation time cost, and the time convolution network adopted by the technical scheme of the patent has sparse characteristics, and the unique expansion convolution structure can greatly reduce the parameter scale of the network on the premise of ensuring the information extraction effect, so that the calculation efficiency is improved;
3. the technical scheme of the invention provides a total fault identification algorithm of a platform area based on a time convolution network (Temporal Convolutional Nets, TCN); the TCN has an expansion convolution structure, and can efficiently extract high-dimensional features in a long sequence and perform fault identification; further, attention mechanism (Attention) is introduced in fault identification of annual sequences to strengthen the screening capability of key information fragments;
4. according to the technical scheme, the fault is identified by extracting the high-dimensional characteristics in the time sequence of the measuring value of the area; further, attention mechanisms are introduced in the annual sequence fault identification so as to screen out key information fragments and improve the identification accuracy; the proposed fault identification algorithm can realize 93.1% of daily fault identification precision and 85.3% of annual fault identification precision, and has higher calculation efficiency compared with the traditional deep learning model.
Drawings
FIG. 1 is a schematic diagram of a time convolution network architecture;
FIG. 2 is a schematic diagram of a multi-headed attention mechanism;
FIGS. 3 a-3 d are exemplary graphs of abnormal power or fault transformer measurements;
FIG. 4 is a schematic diagram of the training process of TCN and CNN-LSTM models;
FIG. 5 is a schematic diagram of TCN model test results;
FIG. 6a is a schematic diagram of an annual sequence fault recognition confusion matrix prior to introduction of an attention mechanism;
FIG. 6b is a schematic diagram of an annual sequence failure recognition confusion matrix after the attention mechanism is turned on;
fig. 7 is a schematic block diagram of a process of intelligent identification method for total fault of a platform area according to the present invention.
Detailed Description
The invention is further described below with reference to the drawings and examples.
1. The modeling method comprises the following steps:
the measurement data of the total table of the transformer area can be expressed as
Where E represents the number of measurement variables (e.g., voltage, current, power, etc.), and L represents the length of the measurement sequence. The transformer failure label is denoted by Y.
In the technical scheme, the transformer fault identification is essentially a classification problem, y=0 represents a normal state, and y=1 represents a fault state. The machine learning of fault identification is expressed as y=f (X), that is, by using the machine learning model f and taking the measurement sequence as input, it is determined whether the transformer has a fault in the measurement time.
1.1 time convolution network:
the time convolution network structure is shown in fig. 1, and is essentially a full convolution network with an expanded structure.
Let the convolution kernel be denoted as f= [ F 0 ,f 1 ,…,f M-1 ]Where M is the number of convolution kernels.
Then for element X in the input matrix X m Is a dilation convolution operation of (a)Can be expressed as:
wherein d is the expansion ratio, and m-d.i is x m Pointers to previous elements.
When d=1, the network then degenerates into a conventional convolutional network.
Whereas when d >1, the dilation convolution operation will skip the (d-1)/d elements in the previous layer and focus on only the remaining 1/d elements.
This feature can significantly improve the receptive field of the network and reduce the complexity of the network.
The receptive field of the model can be calculated by
Where K is the size of the convolution kernel, N stack Is the number of stacks of the network.
According to the calculation formula of the receptive field of the model, the receptive field of the network can be increased by increasing the size of convolution kernels, increasing the stacking number, improving the expansion rate and the like.
Through the network, high-dimensional features in the input data matrix can be extracted and compressed to the output layer and further used for fault identification.
1.2 attentiveness mechanisms:
attention mechanisms are techniques that extract pieces of key information from an input sequence.
The attention function may be described as a mapping between pairs of Query (Q) and Key-Value (K) -Value item (Value, V) data to output, as shown in the following equation:
wherein Q, K, V are vectors and include trainable parameters, and alpha is a scaling factor.
On the basis of using a single attention function, the parallel adoption of multiple attention functions will help to promote the overall performance, i.e. Multi-head attention mechanism (Multi-head attention), as shown in fig. 2.
The calculation method of the multi-head attention mechanism is shown as follows:
MultiHead(Q,K,V)＝concate(head 1 ,head 2 ,...,head n )W O
head i ＝Attention(QW i Q ,KW i K ,VW i V )
wherein W is i Q ，W i K ，W i V And W is O Is a parameter matrix that can be learned during the mapping process.
Based on the attention mechanism, key information fragments in the input sequence can be effectively screened out and are not influenced by the distance between the information fragments.
In summary, as shown in fig. 7, the technical solution of the present invention provides a method for intelligently identifying total table faults of a transformer area based on a time convolution network, which includes collecting and cleaning total table measurement data of each transformer area, and is characterized in that:
1) Establishing a transformer fault label;
2) Parallel measurement time sequence data of different variables are input into a network as a data matrix, and information in the data is extracted in a 2D convolution form;
3) Machine learning of transformer fault identification is performed by adopting a time convolution network;
4) Increasing the receptive field of the network by increasing the size of convolution kernels, increasing the number of stacks and increasing the expansion rate;
5) Extracting and compressing high-dimensional features in an input data matrix to an output layer for further fault identification;
6) On the basis of using a single attention function, a multi-head attention mechanism is adopted to promote the overall performance;
7) Adopting two different fault identification strategies, namely daily fault identification and annual fault identification;
8) Based on a time convolution network, high-dimensional features in a long sequence are efficiently extracted and fault identification is carried out;
9) Attention mechanism is introduced in fault identification of annual sequences to strengthen screening capability of key information fragments
10 By using a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of other unreasonable factors in the lines or the areas is eliminated, so that the economic operation rate of the lines is improved.
Specifically, according to the intelligent identification method for total fault of the platform area based on the time convolution network, measurement time sequence data of different variables are input into the network in parallel as a data matrix, and information in the data matrix is extracted in a 2D convolution mode; extracting time sequence features contained by each variable independently, and simultaneously transversely extracting correlations among the variables; on the basis, the unique expansion convolution structure of the time convolution network is utilized, and the parameter scale of the network is greatly reduced on the premise of guaranteeing the information extraction effect, so that the calculation efficiency is improved.
Furthermore, according to the intelligent identification method for the total fault of the platform area based on the time convolution network, the time convolution network is adopted to efficiently extract the correlations of different measured variables, so that the accuracy of fault identification is improved; and by utilizing a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of lines or other unreasonable factors in the areas is eliminated, so that the economic operation rate of the lines is improved.
2. And (3) carrying out calculation analysis:
the calculation example data used in the technical scheme of the invention is the district measurement data of 2920 10kV transformers in a certain city. The data length was 2019.7.1 to 2020.8.1 for 13 months and the resolution 15 minutes.
Each transformer area measurement value comprises 15 variables in total, namely three-phase voltage, three-phase current, three-phase active power, total active power, reactive power, power factor, voltage unbalance rate, current unbalance rate and load rate.
Of these 2920 transformers, some had a definite charge anomaly signature and a field fault detection signature.
In fig. 3a to 3d, 4 transformers with abnormal power or fault labels are chosen as examples.
The abnormal situation of the station #1 occurs slightly from the beginning of 16 days of 3 months in 2020, and part of missing data exists, but no specific fault label exists, which indicates that the phenomenon may be caused by faults or normal phenomenon caused by the change of the electricity consumption behavior of a user.
There is a significant drop in the three-phase voltage of bay #2 and more missing data. And according to the fault label, the junction box C is displayed as short circuit.
And the platform area #3 is marked as abnormal electric quantity, and the fault is ABC three-phase short circuit. From the measurement data, the available measurement data is very limited.
The cell #4 current has a negative value and the fault is marked as C-phase current head-to-tail on the table.
The above anomaly or fault tags can be used as tags for zone fault identification.
According to the technical scheme, two different fault identification strategies are adopted to verify the validity of the model, namely, daily fault identification and annual fault identification.
The solar fault identification is used for judging whether the transformer is abnormal on the same day by reading a solar measurement value; the annual fault identification reads the measured value of the whole year to determine whether the transformer has faults in the year.
2.1, identifying the abnormality of the degree of the sun:
in the daily degree identification, a training sample set is first established, which includes a normal sample (marked as 0) and an abnormal sample (marked as 1), each sample being a 96×15 data matrix.
Through statistical analysis, the integrity of data of a 10kV transformer area in a certain city is about 80%, and 20% of data are missing.
Therefore, the original data is first preprocessed, the number of days in which the missing data exceeds 10% in each transformer is deleted, and the data complement based on the time-series extrapolation and correlation analysis is performed on the number of days in which the missing data is less than 10%.
On the basis, a total of 364 daily degree samples with clear abnormal labels are further screened out. Meanwhile, 500 daily degree samples are randomly selected from a normally running platform region to form a training set together.
The total 864 training samples were split into training sets (70%, 605), test sets (20%, 173) and validation sets (10%, 65). Training the model by adopting the training set and the verification set, and finally testing the performance of the model on the test set.
It should be noted that in the recognition of solar degree anomalies, because the sequence length is short, the attention mechanism is not adopted, and only the classical TCN structure is used. The model parameter setting of TCN is shown in table 1, and the receptive field is 161 and is greater than the length 96 of the daily degree sequence, so that the information contained in the daily degree sequence can be extracted completely.
TABLE 1TCN daily fail identification model parameter settings
FIG. 4 shows the change in the loss function during training of the TCN model with a convolutional-long short term memory network model (CNN-LSTM) with similar overall parameter scale. It can be seen that the TCN model has a faster convergence rate and test accuracy than the CNN-LSTM model.
Training of the TCN model takes about 3 minutes and becomes substantially stable after 40 iterations.
The TCN model after training is applied to a test set, the classification result confusion matrix is shown in figure 5, and the comprehensive identification accuracy is 93.1%.
2.1, identifying annual faults:
in annual fault identification, the model input is an annual measurement data matrix, with dimensions of [396 x 96,15], which is a typical long-time sequence.
Similar to the daily anomaly identification, 70 transformer bays with clear fault signatures in the target year are first screened out and marked as 1.
Meanwhile, 100 marks are randomly selected from the normal station area to be 0, and a training set is formed together.
For missing data therein, it is uniformly filled with 0.
Similarly, the data set of 170 samples was split into training sets (70%, 119), test sets (20%, 34) and validation sets (10%, 17). For long sequence identification, the critical fault information fragments need to be screened, so that an attention mechanism is introduced. The model parameters are shown in table 2.
TABLE 2TCN annual fault identification model parameter settings
After model training is completed, it is applied to the test set to verify its recognition effect.
To compare the effects of the attention mechanisms, the model results before and after attention mechanism introduction are given as shown in fig. 6.
The comprehensive identification precision before the introduction is 67.6%, and the comprehensive identification precision after the introduction is 85.3%, so that the precision is obviously improved, and particularly the identification precision of a fault sample is obviously improved.
The technical scheme of the invention provides an intelligent identification method for total fault of a platform area based on a time convolution network. The expansion convolution structure of the algorithm can obviously improve the efficiency of the model, and is particularly suitable for extracting the characteristics of long-time sequences. Further, attention mechanisms are introduced in the annual fault identification, so that the extraction capacity of key information fragments can be improved, and the accuracy of the model is improved. The actual calculation result of the 10kV district in a certain city shows that the proposed algorithm can realize 93.1% of daily fault identification precision and 85.3% of annual fault identification precision. The invention can be widely used in the field of electric energy metering management and fault intelligent recognition of the total tables of all the power supply company.
Claims (10)
1. The utility model provides a total table trouble intelligent identification method of district based on time convolution network, includes the collection and the washing of each transformer district total table measurement data, characterized by:
1) Establishing a transformer fault label;
2) Parallel measurement time sequence data of different variables are input into a network as a data matrix, and information in the data is extracted in a 2D convolution form;
3) Machine learning of transformer fault identification is performed by adopting a time convolution network;
4) Increasing the receptive field of the network by increasing the size of convolution kernels, increasing the number of stacks and increasing the expansion rate;
5) Extracting and compressing high-dimensional features in an input data matrix to an output layer for further fault identification;
6) On the basis of using a single attention function, a multi-head attention mechanism is adopted to promote the overall performance;
7) Adopting two different fault identification strategies, namely daily fault identification and annual fault identification;
8) Based on a time convolution network, high-dimensional features in a long sequence are efficiently extracted and fault identification is carried out;
9) Attention mechanism is introduced in fault identification of annual sequences to strengthen screening capability of key information fragments
10 By using a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of other unreasonable factors in the lines or the areas is eliminated, so that the economic operation rate of the lines is improved.
2. The intelligent identification method for total fault of transformer area based on time convolution network as set forth in claim 1, wherein said total measurement data of transformer area is expressed as:
wherein E represents the number of measurement variables and L represents the length of the measurement sequence.
3. The intelligent recognition method for total fault of a platform area based on a time convolution network according to claim 2, wherein the measurement variables at least comprise voltage, current and power.
4. The intelligent identification method for total fault in a transformer area based on a time convolution network according to claim 1, wherein the transformer fault is identified as a classification problem, y=0 represents a normal state, and y=1 represents a fault state;
the machine learning of fault identification is expressed as y=f (X), that is, by using the machine learning model f and taking the measurement sequence as input, it is determined whether the transformer has a fault in the measurement time.
5. The intelligent identification method for total fault of a district based on a time convolution network according to claim 1, wherein the time convolution network is a full convolution network with an expansion structure;
let the convolution kernel be denoted as f= [ F 0 ,f 1 ,…,f M-1 ]Wherein M is the number of convolution kernels;
then for element X in the input matrix X m Is a dilation convolution operation of (a)Expressed as:
wherein d is the expansion ratio, and m-d.i is x m A pointer to a previous element;
when the expansion ratio d=1, the network is degraded into a conventional convolutional network;
when the expansion ratio d >1, the expansion convolution operation skips the element of (d-1)/d in the previous layer and focuses on the rest 1/d elements only;
the above characteristics of the expansion ratio can significantly improve the receptive field of the network and reduce the complexity of the network.
6. The intelligent recognition method for total fault of a platform area based on a time convolution network according to claim 5, wherein the receptive field is calculated by the following formula:
where K is the size of the convolution kernel, N stack Is the number of stacks of the network.
7. The intelligent recognition method for the total table faults of the area based on the time convolution network according to claim 1 is characterized in that the intelligent recognition method for the total table faults of the area based on the time convolution network is characterized in that on the basis of using a single attention function, a plurality of attention functions are adopted in parallel to help to promote the overall performance, namely a multi-head attention mechanism;
the calculation of the multi-head attention mechanism is shown as follows:
MultiHead(Q,K,V)＝concate(head 1 ,head 2 ,...,head n )W O
head i ＝Attention(QW i Q ,KW i K ,VW i V )
wherein W is i Q ，W i K ，W i V And W is O Is a parameter matrix that can be learned during the mapping process.
8. The intelligent identification method for total fault of the area based on the time convolution network according to claim 1, wherein the intelligent identification method for total fault of the area based on the time convolution network uses a classical TCN structure in the identification of the daily degree anomaly; in annual fault identification, attention mechanisms are introduced to improve the extraction capacity for key information pieces, thereby improving the accuracy of the model.
9. The intelligent identification method for total fault of the area based on the time convolution network according to claim 1 is characterized in that the intelligent identification method for total fault of the area based on the time convolution network takes measurement time sequence data of different variables in parallel as a data matrix to be input into the network, and extracts information in the data matrix in a form of 2D convolution; extracting time sequence features contained by each variable independently, and simultaneously transversely extracting correlations among the variables; on the basis, the unique expansion convolution structure of the time convolution network is utilized, and the parameter scale of the network is greatly reduced on the premise of guaranteeing the information extraction effect, so that the calculation efficiency is improved.
10. The intelligent identification method for the total fault of the area based on the time convolution network is characterized in that the intelligent identification method for the total fault of the area based on the time convolution network adopts the time convolution network to efficiently extract the correlations of different measured variables so as to improve the accuracy of fault identification; and by utilizing a deep learning tool, the monitoring data of the areas are automatically and batched analyzed, the total table of the areas with faults is directly positioned, and the interference of lines or other unreasonable factors in the areas is eliminated, so that the economic operation rate of the lines is improved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310636611.4A CN116879672A (en) | 2023-05-31 | 2023-05-31 | Intelligent identification method for total fault of area based on time convolution network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310636611.4A CN116879672A (en) | 2023-05-31 | 2023-05-31 | Intelligent identification method for total fault of area based on time convolution network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116879672A true CN116879672A (en) | 2023-10-13 |
Family
ID=88259303
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310636611.4A Pending CN116879672A (en) | 2023-05-31 | 2023-05-31 | Intelligent identification method for total fault of area based on time convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116879672A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725491A (en) * | 2024-02-18 | 2024-03-19 | 四川大学 | SCITET-based power system fault state detection and classification method |
-
2023
- 2023-05-31 CN CN202310636611.4A patent/CN116879672A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117725491A (en) * | 2024-02-18 | 2024-03-19 | 四川大学 | SCITET-based power system fault state detection and classification method |
| CN117725491B (en) * | 2024-02-18 | 2024-05-03 | 四川大学 | SCINet-based power system fault state detection and classification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106779505B (en) | Power transmission line fault early warning method and system based on big data driving | |
| CN111584952A (en) | Method and system for online evaluation of electrochemical cells of energy storage power station | |
| CN108664010A (en) | Generating set fault data prediction technique, device and computer equipment | |
| CN110458230A (en) | A kind of distribution transforming based on the fusion of more criterions is with adopting data exception discriminating method | |
| CN108649892B (en) | Defect diagnosis method and device for photovoltaic power station | |
| CN111950165B (en) | Electric energy meter reliability analysis method under typical environment | |
| CN116879672A (en) | Intelligent identification method for total fault of area based on time convolution network | |
| CN115469227A (en) | Set variational self-encoder and dynamic regular lithium battery abnormity detection method | |
| CN116401585B (en) | Energy storage battery failure risk assessment method based on big data | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| Yun et al. | Research on fault diagnosis of photovoltaic array based on random forest algorithm | |
| Yoon et al. | Deep learning-based method for the robust and efficient fault diagnosis in the electric power system | |
| CN116466241A (en) | Thermal runaway positioning method for single battery | |
| CN115128468A (en) | Chemical energy storage battery PHM undervoltage fault prediction method | |
| CN111626559A (en) | Main factor analysis method-based medium-voltage distribution network line loss key characteristic index extraction method and system | |
| CN112926686B (en) | BRB and LSTM model-based power consumption anomaly detection method and device for big power data | |
| CN116933201A (en) | Method and system for identifying illegal electricity utilization behavior of low-voltage charging pile | |
| CN117493923A (en) | Method and system for repairing abnormal data of low-voltage distribution transformer area containing distributed photovoltaic | |
| CN108802565B (en) | Medium-voltage power distribution network disconnection ungrounded fault detection method based on machine learning | |
| CN115267586B (en) | SOH (solid oxide fuel cell) evaluation method of lithium battery | |
| CN113866562B (en) | Power system transformer area line loss identification method | |
| CN117764547A (en) | Photovoltaic string fault diagnosis method and system | |
| CN113961708B (en) | Power equipment fault tracing method based on multi-level graph convolutional network | |
| CN112256735B (en) | Power consumption monitoring method and device, computer equipment and storage medium | |
| CN116879671A (en) | District fault identification method based on time convolution network and attention mechanism |
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 |