CN114444375A - High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation - Google Patents
High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation Download PDFInfo
- Publication number
- CN114444375A CN114444375A CN202111522095.XA CN202111522095A CN114444375A CN 114444375 A CN114444375 A CN 114444375A CN 202111522095 A CN202111522095 A CN 202111522095A CN 114444375 A CN114444375 A CN 114444375A
- Authority
- CN
- China
- Prior art keywords
- data
- switch cabinet
- classification
- sensor
- value
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/04—Ageing analysis or optimisation against ageing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention discloses a method for realizing the method, which is realized by the following technical scheme: the method comprises the steps of obtaining sensor data in the switch cabinet, preprocessing the data, constructing and training an insulation level prediction model based on LSTM, fusing the parameter classification models of each sensor, evaluating the classification capability of 5 LSTM classifiers by adopting a validation sample, combining all the LSTM classifiers into a strong classifier, and predicting the insulation level of the switch cabinet at the moment according to the monitoring result of the sensor. The method adopts an artificial intelligence technology, takes a circulating neural network as a core, preprocesses monitoring information, selects a reasonable network structure and adopts a classification method to realize the grade prediction of the insulation aging degree of the switch cabinet. According to the network input and output requirements, cleaning, dimension reduction and feature extraction are carried out on the multi-parameter time-varying information, an effective data structure is formulated, and training of data resources on the model is completed.
Description
Technical Field
The invention relates to a high-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation.
Background
The high-voltage switch cabinet is widely applied to a power transformation and distribution system, and plays a role in controlling and protecting a circuit. Once the high-voltage switch cabinet breaks down, a large-area power failure accident can be caused, and great loss is caused for production and life. Because the high tension switchgear structure is complicated, the trouble is of a great variety. The insulation performance deterioration is the main reason for causing the cubical switchboard trouble, however the judgement degree of difficulty of insulating state is big, and conventional preventive insulation experiment can not cover the cubical switchboard and move the full cycle, is difficult to ensure the safe operation of cubical switchboard.
The reasons for inducing insulation aging and failure are many, and mainly include insulation degradation caused by external factors and the quality of the equipment itself.
(1) Influence of environmental conditions 1) over-high or over-low temperature, when the switch is operated in an over-high temperature environment for a long time, the insulation material of the switch is easy to age, so that the insulation level is reduced; 2) the dirt deposited on the surface of the insulation not only influences the power frequency voltage-withstand characteristic of the insulation, but also greatly reduces the impact performance of the insulation. 3) And when the electric equipment is wet, the electrolyte of the polluted layer is gradually decomposed after being wetted, and a thin conductive liquid film is formed on the insulating surface, so that the insulating strength of the electric equipment is greatly reduced.
(2) The equipment is poor in manufacturing quality and process, and the manufacturing quality and the assembling quality have great influence on the whole voltage-resistant level of the switch cabinet, such as irregular screws, poor assembling quality, poor quality of supporting porcelain columns or porcelain sleeves, poor stability and breakage under short-circuit current impact; 2) the contact capacity is insufficient or the contact is poor, resulting in local heat generation. 3) Insufficient creepage and air gap are the root causes of insulation damage accidents of the switch cabinet. In particular to a handcart cabinet, in order to shorten the size of the cabinet body, production units often greatly reduce the distance between circuit breakers and isolating plugs or the distance to the ground, and no effective measure for ensuring the insulating strength is taken.
(3) The equipment running time is too long, so that the insulating material is naturally aged, and the insulating property is reduced.
In addition, lightning overvoltage and overvoltage impact in the using process can also accelerate the aging speed of the insulating material.
Based on the above insulation aging reasons, various insulation detection methods have appeared, which mainly include: partial discharge detection (including transient earth voltage TEV monitoring, ultrasonic detection and ultrahigh frequency intelligent sensor detection), and operation environment monitoring (temperature and humidity monitoring). Although the monitoring methods reflect the insulation aging state to a certain extent, the monitoring results are given independently at present, the monitoring methods depend on manual judgment, fusion calculation and intelligent analysis of multiple parameters are lacked, and the reliability is influenced.
Disclosure of Invention
The invention aims to provide a high-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation, which can effectively solve the problems that the existing switch cabinet monitoring method mostly gives a monitoring result independently, depends on artificial judgment, lacks multi-parameter fusion calculation and intelligent analysis, and has influenced reliability.
In order to solve the technical problems, the invention is realized by the following technical scheme: the high-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation comprises the following steps:
s10: acquiring data of a sensor in a switch cabinet;
s20: preprocessing the data;
s30: constructing and training an insulation level prediction model based on LSTM: including the steps S31 and S32,
s31: constructing a classification model based on an LSTM, wherein the LSTM network comprises 24 cells, the data input of each cell is a 10x1 vector, the output of a hidden layer is a 32x1 vector, the output of a full connection layer is a 5x1 vector, and 5 levels of probabilities are output through a softmax function;
s32: importing the training samples into a network, and training a classification model for each sensor parameter;
s40: fusing various sensor parameter classification models: including step S41 to step S43:
s41: evaluating the respective classification capability of the 5 LSTM classifiers by adopting an examination sample; let the kth sample label class be j, define
Relative value being the difference between the correct class attribute and the maximum value of the incorrect class attribute
Wherein the content of the first and second substances,
representing the probability of j in the test output result of the kth sample when the ith classifier is adopted;
it is indicated that the classification is correct,
the larger, the lower the risk of error; classification Capacity of classifier i on all samples in class j Using all samples in class j
Average value of (2)
Represents:
s42: combining all LSTM classifiers into a strong classifier; for any test sample, the attribute value of the LSTM classifier i for dividing the test sample into j classes is set as
Normalization value of classification capability of combined classifier i on j type sample
Obtaining the probability that the sample belongs to the j class
S43: outputting a classification result: for the input sample, take PjJ corresponding to the maximum value of j 1.. and 5 is used as a classification result, namely the strong classifier considers that the insulation level of the switch cabinet at the moment is j predicted according to the monitoring result of the sensor.
Preferably, the step S10 of acquiring the sensor data in the switch cabinet includes:
s11: the method comprises the following steps of installing an ultrasonic transient ground voltage temperature all-in-one sensor and an ultrahigh frequency sensor for a switch cabinet, installing a temperature sensor and a humidity sensor in a switching station, and acquiring data of partial discharge failure information and environmental condition information in the operation process of the switch cabinet;
s12: and (4) periodically testing the insulation level of the switch cabinet and recording data.
Preferably, the step S20 of preprocessing the data includes:
s21: taking data 24 hours before the insulation level test as training data, and taking the recorded insulation level as a prediction result;
s22: dividing the training data obtained in step S21 into 24 groups, one group per hour, and arranging in chronological order;
s23: and extracting characteristics of each group of data.
Preferably, the step S23 of extracting features from each group of data includes the following steps:
s231: carrying out median filtering to remove noise interference;
s232: dividing the 1-hour data according to the time of 30 minutes, 15 minutes, 8 minutes, 4 minutes, 2 minutes and 1 minute, and then calculating the average value to obtain 6 average values;
s233: calculating the maximum positive jump value and the maximum negative jump value within 1 hour to obtain two jump values;
s234: calculating the mean value and the variance within 1 hour to obtain two values;
s235: the above values are combined in sequence to obtain 10 values, the values of all sensors are processed according to the same method to obtain training data, the data are divided into a training group, a validation group and a test group, and the recorded insulation level is a classification label value.
Compared with the prior art, the invention has the advantages that: the method adopts an artificial intelligence technology, takes a circulating neural network as a core, preprocesses monitoring information, selects a reasonable network structure and adopts a classification method to realize the grade prediction of the insulation aging degree of the switch cabinet. According to the network input and output requirements, cleaning, dimension reduction and feature extraction are carried out on the multi-parameter time-varying information, an effective data structure is formulated, and training of data resources on the model is completed. And a reasonable optimization method is adopted, the training and recognition speed is increased, and the insulation degree grade prediction accuracy is improved.
Drawings
FIG. 1 is a diagram of a recurrent neural network architecture to be employed in the present invention;
FIG. 2 is a diagram of a classification model of a single monitor according to the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
Referring to fig. 1 and fig. 2, the embodiment of the method for predicting insulation aging of a high-voltage switch cabinet based on multi-parameter fusion calculation according to the present invention, the method for predicting insulation aging of a high-voltage switch cabinet based on multi-parameter fusion calculation includes the following steps:
s10: acquiring sensor data in the switch cabinet, including steps S11 and S12;
s11: the method comprises the steps that an ultrasonic transient ground voltage temperature all-in-one sensor and an ultrahigh frequency sensor are installed on a switch cabinet, and a temperature sensor and a humidity sensor are installed in a switching station and used for collecting data of partial discharge failure information and environmental condition information in the operation process of the switch cabinet;
s12: testing and recording the insulation level of the switch cabinet every 3 months according to the insulation level test requirements of the switch cabinet specified by the national standard, and testing for 4 times;
s20: preprocessing the data, including steps S21 to S23:
s21: taking data 24 hours before the insulation level test as training data, and taking the recorded insulation level as a prediction result;
s22: dividing the 24-hour data into 24 groups, one group per hour, in chronological order;
s23: extracting features for each set of data, including steps S231 to S235:
s231: performing median filtering to remove noise interference;
s232: dividing the 1-hour data according to the time of 30 minutes, 15 minutes, 8 minutes, 4 minutes, 2 minutes and 1 minute, and then calculating the average value to obtain 6 average values;
s233: calculating the maximum positive jump value and the maximum negative jump value within 1 hour to obtain two jump values;
s234: calculating the mean value and the variance within 1 hour to obtain two values;
s235: the above values were combined in order to obtain 10 values. Processing the values of all 5 sensors according to the same method to obtain training data, namely, the detection value of each sensor is 24 groups, and each group comprises 10 sensors;
as much data as possible was collected, the data was divided into training, validation and test groups, and the insulation levels recorded were the classification label values.
S30: the LSTM-based insulation level prediction model is constructed and trained, including steps S31 and S32:
s31: constructing a classification model based on an LSTM (Long Short-Term Memory network), as shown in FIG. 2, the LSTM network comprises 24 cells, the data input of each cell is a 10x1 vector, the output of a hidden layer is a 32x1 vector, the output of a full connection layer is a 5x1 vector, and 5 levels of probabilities are output through a softmax function;
s32: importing the training samples into a network, and training a classification model for each sensor parameter;
s40: fusing various sensor parameter classification models: including step S41 to step S43:
s41: evaluating the respective classification capability of the 5 LSTM classifiers by adopting an examination sample; let the kth sample label category be j, define
Relative value being the difference between the correct class attribute and the maximum value of the incorrect class attribute
Wherein, the first and the second end of the pipe are connected with each other,
representing the probability of j in the test output result of the kth sample when the ith classifier is adopted;
represents the classification ofIndeed, the content of the active ingredient is,
the larger, the lower the risk of error; classification Capacity of classifier i on all samples in class j Using all samples in class j
Average value of (2)
Represents:
s42: combining all LSTM classifiers into a strong classifier; for any test sample, the attribute value of the LSTM classifier i for dividing the test sample into j classes is set as
Normalization value of classification capability of combined classifier i on j type sample
Obtaining the probability that the sample belongs to the j class
S43: outputting a classification result: for the input sample, take PjJ corresponding to the maximum value of j 1.. and 5 is taken as a classification result, namely the strong classifier considers that the insulation level of the switch cabinet at the moment is j predicted according to the monitoring results of all 5 sensors in the past 24 hours;
s50: and (4) corresponding the insulation level of the switch cabinet at the moment to the measured values of 4 times.
According to the invention, a prediction model is established by using multi-parameter historical data and faults, the insulation aging degree is predicted according to real-time online monitoring information, the safety is improved, and the accident risk is reduced. The method comprises the steps of collecting and fusing multi-source information monitored by the switch cabinet, establishing a network model by using an artificial intelligence method, combining information such as partial discharge, operating environment, operating time and the like, completing feature extraction and classification identification, realizing prediction of the insulation aging degree of the switch cabinet, and providing reliable guidance and early warning information for operation and maintenance of the high-voltage switch cabinet.
The above description is only an embodiment of the present invention, but the technical features of the present invention are not limited thereto, and any changes or modifications within the technical field of the present invention by those skilled in the art are covered by the claims of the present invention.
Claims (4)
1. The high-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation is characterized by comprising the following steps: the method comprises the following steps:
s10: acquiring data of a sensor in a switch cabinet;
s20: preprocessing the data;
s30: constructing and training an insulation level prediction model based on LSTM: including the steps S31 and S32,
s31: constructing a classification model based on an LSTM, wherein the LSTM network comprises 24 cells, the data input of each cell is a 10x1 vector, the output of a hidden layer is a 32x1 vector, the output of a full connection layer is a 5x1 vector, and 5 levels of probabilities are output through a softmax function;
s32: importing the training samples into a network, and training a classification model for each sensor parameter;
s40: fusing various sensor parameter classification models: including step S41 to step S43:
s41: evaluating the respective classification capability of the 5 LSTM classifiers by adopting an examination sample; let the kth sample label class be j, define
Relative value being the difference between the correct class attribute and the maximum value of the incorrect class attribute
Wherein the content of the first and second substances,
representing the probability of j in the test output result of the kth sample when the ith classifier is adopted;
it is indicated that the classification is correct,
the larger, the lower the risk of error; classification Capacity of classifier i on all samples in class j Using all samples in class j
Average value of (2)
Represents:
s42: combining all LSTM classifiers into a strong classifier; for any test sample, the attribute value of the LSTM classifier i for dividing the test sample into j classes is set as
Normalization value of classification capability of combined classifier i on j type sample
Obtaining the probability that the sample belongs to the j class
S43: outputting a classification result: for the input sample, take PjJ corresponding to the maximum value of j 1.. and 5 is used as a classification result, namely the strong classifier considers that the insulation level of the switch cabinet at the moment is j predicted according to the monitoring result of the sensor.
2. The method for predicting the insulation aging of the high-voltage switch cabinet based on the multi-parameter fusion calculation as claimed in claim 1, wherein: the step S10 of acquiring the sensor data in the switch cabinet includes:
s11: the method comprises the following steps of installing an ultrasonic transient ground voltage temperature all-in-one sensor and an ultrahigh frequency sensor for a switch cabinet, installing a temperature sensor and a humidity sensor in a switching station, and acquiring data of partial discharge failure information and environmental condition information in the operation process of the switch cabinet;
s12: and (4) periodically testing the insulation level of the switch cabinet and recording data.
3. The method for predicting the insulation aging of the high-voltage switch cabinet based on the multi-parameter fusion calculation as claimed in claim 1, wherein: the specific step of preprocessing the data in step S20 includes:
s21: taking data 24 hours before the insulation level test as training data, and taking the recorded insulation level as a prediction result;
s22: dividing the training data obtained in step S21 into 24 groups, one group per hour, and arranging in chronological order;
s23: and extracting characteristics of each group of data.
4. The method for predicting the insulation aging of the high-voltage switch cabinet based on the multi-parameter fusion calculation as claimed in claim 3, wherein: the step S23 of extracting features for each set of data includes the steps of:
s231: carrying out median filtering to remove noise interference;
s232: dividing the 1-hour data according to the time of 30 minutes, 15 minutes, 8 minutes, 4 minutes, 2 minutes and 1 minute, and then calculating the average value to obtain 6 average values;
s233: calculating the maximum positive jump value and the maximum negative jump value within 1 hour to obtain two jump values;
s234: calculating the mean value and the variance within 1 hour to obtain two values;
s235: the above values are combined in sequence to obtain 10 values, the values of all sensors are processed according to the same method to obtain training data, the data are divided into a training group, a validation group and a test group, and the recorded insulation level is a classification label value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111522095.XA CN114444375A (en) | 2021-12-13 | 2021-12-13 | High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111522095.XA CN114444375A (en) | 2021-12-13 | 2021-12-13 | High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114444375A true CN114444375A (en) | 2022-05-06 |
Family
ID=81364029
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111522095.XA Pending CN114444375A (en) | 2021-12-13 | 2021-12-13 | High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114444375A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117891160A (en) * | 2024-03-13 | 2024-04-16 | 陕西西高电气科技有限公司 | Intelligent control system and method for switch cabinet |
| CN117891160B (en) * | 2024-03-13 | 2024-05-31 | 陕西西高电气科技有限公司 | Intelligent control system and method for switch cabinet |
-
2021
- 2021-12-13 CN CN202111522095.XA patent/CN114444375A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117891160A (en) * | 2024-03-13 | 2024-04-16 | 陕西西高电气科技有限公司 | Intelligent control system and method for switch cabinet |
| CN117891160B (en) * | 2024-03-13 | 2024-05-31 | 陕西西高电气科技有限公司 | Intelligent control system and method for switch cabinet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106199305A (en) | Underground coal mine electric power system dry-type transformer insulation health state evaluation method | |
| CN105866638A (en) | City network cable joint insulation state online monitoring apparatus early warning apparatus and method | |
| CN111025097B (en) | Intelligent method for evaluating leakage current difference factor of XLPE cable after aging | |
| CN112152313B (en) | Method for identifying power equipment by acquisition system | |
| CN108459269A (en) | A kind of 10kV pvs (pole-mounted vacuum switch)s state on-line evaluation method and apparatus | |
| CN113268590A (en) | Power grid equipment running state evaluation method based on equipment portrait and integrated learning | |
| CN109597396A (en) | A kind of distribution transforming on-line fault diagnosis method based on high amount of traffic and transfer learning | |
| CN116845391A (en) | Lithium battery energy storage management system | |
| CN113486291B (en) | Petroleum drilling machine micro-grid fault prediction method based on deep learning | |
| CN117390403B (en) | Power grid fault detection method and system for new energy lighthouse power station | |
| CN114740306A (en) | Power grid informatization-based power distribution network line fault online monitoring and early warning management system | |
| CN113791324A (en) | Pollution insulator flashover risk early warning system and method | |
| CN114444375A (en) | High-voltage switch cabinet insulation aging prediction method based on multi-parameter fusion calculation | |
| CN111999695A (en) | State evaluation and abnormity diagnosis method for metering device of transformer substation | |
| GU et al. | Research on intelligent early warning algorithm for distribution network considering extreme climate conditions | |
| CN117009731A (en) | Power failure early warning method based on meteorological factors | |
| CN116840631A (en) | Transformer partial discharge monitoring and positioning method based on joint diagnosis model | |
| CN116644361A (en) | Overhead transmission line fault diagnosis method and computer readable medium | |
| CN103513168B (en) | GIS and cable local discharge comprehensive judging method | |
| CN115792634A (en) | Battery monomer voltage sampling fault identification method based on cloud online data | |
| CN112904226B (en) | Method for rapidly judging short-circuit fault of high-voltage bus based on induced electricity | |
| CN114997375A (en) | Insulator surface metal pollution model building system in air type switch cabinet | |
| CN115169932A (en) | Comprehensive evaluation method for state of high-voltage circuit breaker | |
| Wu et al. | XLPE cable health assessment based on Relief-F feature weighted FSVM | |
| CN115224684A (en) | Intelligent power distribution network risk state identification method and system based on immune hazard theory |
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 |