CN115561596A - Lightning arrester insulation state assessment method based on random forest - Google Patents
Lightning arrester insulation state assessment method based on random forest Download PDFInfo
- Publication number
- CN115561596A CN115561596A CN202211268533.9A CN202211268533A CN115561596A CN 115561596 A CN115561596 A CN 115561596A CN 202211268533 A CN202211268533 A CN 202211268533A CN 115561596 A CN115561596 A CN 115561596A
- Authority
- CN
- China
- Prior art keywords
- data
- random forest
- training
- insulation state
- model
- 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
- 238000009413 insulation Methods 0.000 title claims abstract description 40
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims description 34
- 238000001514 detection method Methods 0.000 claims abstract description 18
- 238000011156 evaluation Methods 0.000 claims abstract description 14
- 239000000463 material Substances 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 28
- 238000012360 testing method Methods 0.000 claims description 25
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000003066 decision tree Methods 0.000 claims description 10
- 238000012544 monitoring process Methods 0.000 claims description 9
- 238000013210 evaluation model Methods 0.000 claims description 8
- 238000002790 cross-validation Methods 0.000 claims description 7
- 238000002759 z-score normalization Methods 0.000 claims description 6
- 230000032683 aging Effects 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000009422 external insulation Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013016 damping Methods 0.000 claims description 2
- 238000012706 support-vector machine Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 238000009421 internal insulation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
Images
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
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention relates to the technical field of power grid material quality detection, in particular to a lightning arrester insulation state evaluation method based on random forest.
Description
Technical Field
The invention relates to the technical field of power grid material quality detection, in particular to a lightning arrester insulation state evaluation method based on random forests.
Background
The lightning arrester is an important overvoltage protection device in a power grid, avoids the protected equipment from being damaged by lightning overvoltage or operation overvoltage, and is powerful guarantee for safe operation of a power system. The lightning arrester has the defects of internal insulation, damping, valve block aging and the like during operation, and the safe operation of electric power is influenced. In order to ensure the stable operation of the power grid and the safety of protected equipment and prevent the lightning arrester from causing more serious power grid accidents due to self faults, a preventive test is required to be carried out periodically to check whether the working state of the lightning arrester is good or not.
At present, the method of periodic test, live detection and on-line monitoring is widely used in the power grid. The protected equipment needs to be shut down in a regular test to avoid the damage of the insulation of the protected equipment, so the test condition is usually not the condition for the real operation of the lightning arrester, the accuracy of the test result cannot be ensured, and the method for the regular test cannot prevent the fault occurring in the interval time of the two tests.
The method for live detection and online monitoring is based on the total leakage current, the resistive leakage current and the discharge frequency of the lightning arrester, and judges whether the lightning arrester has a fault or not according to the condition that whether the current value is abnormally increased or not, so that the fault state of the lightning arrester can be reflected to a certain degree theoretically. However, because the actual working and operating environment of the lightning arrester is complex, the accuracy of the measured data can be greatly reduced due to the interference of environmental factors such as temperature, humidity and surface dirt, the single measured value is difficult to be used as the basis for judging the insulation state of the lightning arrester, and the internal operating state of the lightning arrester can be accurately analyzed by combining other related variables.
Based on the reasons, the invention designs an arrester insulation state assessment method based on a random forest, which comprises the steps of selecting characteristic parameters from live detection information, on-line monitoring information, material detection information and environment information of an arrester to form an insulation state characteristic variable data set, constructing an arrester insulation state assessment model based on a random forest algorithm, and providing theoretical support for operation and maintenance of a power grid arrester according to an assessment result so as to exert the lightning protection function of the arrester to the maximum extent.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a lightning arrester insulation state evaluation method based on a random forest.
In order to achieve the purpose, the invention provides a lightning arrester insulation state assessment method based on a random forest, which comprises the following steps:
s1, acquiring samples to be tested of the lightning arrester in different insulation state types, wherein the insulation state types of the lightning arrester can be roughly divided into four types, namely internal damp, insulation aging, external insulation dirt and other types according to historical fault data samples and field actual operation experience;
s2, the characteristic data of the lightning arrester mainly comprise live detection information, online monitoring information, material detection information and environment information;
and S3, carrying out data normalization processing on the data set obtained in the S2 by adopting a z-score normalization method, wherein the calculation formula is as follows:
the average value of new data obtained after the z-score normalization treatment is 0, and the standard deviation is 1;
s4, dividing the normalized data, wherein a part of data is used as a test set, and a part of data is used as a training set;
s5, constructing an arrester insulation state evaluation model on the training set by using a random forest algorithm, and simultaneously obtaining two key parameters of the random forest by using a Grid Search (Grid Search) and cross validation mode: the optimal values of the number of features per tree (Mfeature) and the number of trees (Ntree) enable the stochastic tree model to better fit the training samples;
the cross validation method in S5 comprises the following steps:
randomly dividing a training sample into m mutually disjoint subsets S1, S2, S3, \8230;. Sm, wherein the size of each subset is approximately equal, m-1 parts are selected as a training set each time, and the rest 1 part is used as a test set for training and testing; training and testing are carried out for m times, each sub-sample is tested once, different test effects for m times are obtained, and the average result for m times is used as the final model effect;
the grid searching method in S5 comprises the following steps:
the random forest method has two key parameters, the effectiveness of the model is greatly influenced by the value of the two key parameters, namely the characteristic quantity (Mfeature) of each tree and the number (Ntree) of the trees;
setting value intervals of the feature quantity Mfeature and the number Ntree of the trees, setting corresponding step length, traversing all parameter combinations, combining cross validation, calculating the size of the area (ROC _ auc) under an evaluation index ROC curve of each combination, and determining the optimal feature quantity Mfeature and the number Ntree of the trees, thereby obtaining an optimal arrester insulation state classification prediction model;
s6, based on the test set data in the S5, evaluating the model obtained by training by using indexes such as overall accuracy, recall ratio, precision ratio and the like;
and S7, if the evaluation index of S6 meets the requirement, inputting the sample to be tested in S1 into the lightning arrester insulation state evaluation model, and obtaining the insulation state type of the sample to be tested.
The random forest algorithm comprises the following steps:
s51, selecting a sample from the original data set as a training set by adopting a sampling method (bootstrapping) with a feedback function in the training data set;
s52, generating a decision tree by using the training set obtained by sampling, wherein when each node of the tree is generated:
the k features are selected randomly and non-repetitively, typically
M is the number of characteristic variables;
the k characteristics are used for dividing the sample set respectively to find the optimal division characteristics, and the optimal division characteristics can be distinguished by using a Gini coefficient, a gain rate and an information gain index;
s53, repeating S51-S52 for K times, wherein K is the number of decision trees in the random forest;
and S54, establishing a random forest classification model, and obtaining a final prediction result by adopting a majority voting method because the state evaluation is a classification problem, namely voting the prediction results of all decision trees in the random forest model, wherein the highest obtained class is the final output class.
Compared with the prior art, the lightning arrester insulation state evaluation method has the advantages that the insulation state characteristic variable data set is formed by selecting the characteristic parameters from the live detection information, the on-line monitoring information, the material detection information and the environment information of the lightning arrester, the lightning arrester insulation state evaluation model is constructed based on the random forest algorithm, the evaluation result can provide theoretical support for the operation and maintenance of the power grid lightning arrester, and the lightning protection function of the lightning arrester is exerted to the greatest extent.
Drawings
FIG. 1 is a schematic diagram of a random forest algorithm construction process of the present invention.
Fig. 2 is a flow chart of the arrester insulation state evaluation according to the present invention.
FIG. 3 is a schematic diagram of the random forest parameter optimizing process of the present invention.
Detailed Description
The invention will now be further described with reference to the accompanying drawings.
Referring to fig. 1 to 3, the invention provides a lightning arrester insulation state assessment method based on random forests, comprising the following steps:
1. and acquiring lightning arrester samples with different insulation state types. According to historical fault data samples and field actual operation experience, the insulation state types of the lightning arrester can be roughly classified into four types: internal moisture, insulation aging, external insulation dirt, and other types. 3670 pieces of recorded data of the lightning arresters in the transformer station of 2018-2021 are selected. The category C = { normal, abnormal-aged, abnormal-damped, abnormal-surface dirt, abnormal-other reasons } included in the sample, wherein the arrester abnormal-aged sample proportion is 6%, the abnormal-damped sample proportion is 10%, the abnormal-surface dirt sample proportion is 8%, and the abnormal-other reasons sample proportion is 5%.
2. The characteristic data of the lightning arrester mainly comprises live detection information, on-line monitoring information, material detection information and environment information. The material detection data mainly comprise standing book data, a delivery test report, a handover test report and the like. The invention selects representative characteristic variables which can effectively reflect the operation condition of the arrester, and the specific table is as follows:
| data sources | Type of data |
| On-line monitoring data | Leakage current, resistive current, number of operations |
| Charged detection data | Infrared, partial discharge |
| Material detection | Leakage current, resistive current |
| Environmental data | Temperature, humidity, weather |
3. And (3) carrying out data normalization processing on the data set obtained in the step (2), wherein a z-score normalization method is adopted, and the calculation formula is as follows:
wherein
The new data obtained after the Z-Score normalization process had an average value of 0 and a standard deviation of 1.
4. The normalized data was segmented, with 80% of the data (2936) as the test set and 20% of the data (734) as the training set.
5. Parameter optimization is carried out by adopting a grid searching and cross validation mode, the value intervals of parameters Mfeature and Ntree of the random forest are set as [5,10] and [50,300], corresponding step lengths mstep =1 and nstep =50 are set, and the parameters are combined pairwise to form all possible parameter combination candidate sets, as shown in the following table:
the random number of the training sample is 3, the optimization process is shown in fig. 3, and based on the AUC value of the evaluation index ROC curve, the optimization results are Mfeature =150 and Ntree =4. And constructing a random forest model based on the two optimal parameters to obtain an arrester insulation state evaluation model.
6. And evaluating the model obtained by training based on the data of the test set by using indexes such as overall accuracy, recall ratio, precision ratio and the like. The prediction results of the respective insulation states are shown in the following table:
| valuation index | Is normal | Aging of | Affected with damp | Surface contamination | Other reasons |
| Precision ratio | 93.4％ | 84.3％ | 83.2％ | 80.1％ | 82.1％ |
| Recall ratio of | 92.5％ | 85.5％ | 78.51 | 79.2％ | 80.6％ |
It can be seen from the above table that the recall ratio and precision ratio of "normal" are the highest, the precision ratio and precision ratio of other fault types exceed 78%, and the model has higher prediction capability for each fault type. In addition, the average prediction accuracy of the model was 88.6%.
In order to further verify the superiority of the random forest model, the algorithm is compared with a model constructed by a decision tree and a support vector machine, the comparison process is repeated for 10 times, 80% of data are randomly extracted from original data as training samples each time, the random forest, the decision tree and the support vector machine are constructed by using the training samples, the rest 20% of data are used as test samples, and the prediction accuracy of different models is calculated. The average values of the overall accuracy of the random forest model, the decision tree model and the support vector machine model are 89.12%,85.23% and 84.75%, respectively. Therefore, the generalization capability of the random forest model is better, and the prediction accuracy is higher.
7. And inputting the sample to be tested into the arrester insulation state evaluation model to obtain the insulation state type of the sample to be tested.
The above are only preferred embodiments of the present invention, and are only used to help understanding the method and the core idea of the present application, the scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the scope of the present invention. It should be noted that modifications and adaptations to those skilled in the art without departing from the principles of the present invention should also be considered as within the scope of the present invention.
The lightning arrester insulation state evaluation model based on the random forest algorithm solves the technical problems that in the prior art, the measured quantity is difficult to serve as a judgment basis of a final result due to the influence of uncertain factors such as external temperature, humidity and pollution conditions, combines multiple information such as lightning arrester live line detection and online monitoring, considers the most typical characteristic variable representing the operation state, and has high reliability. Compared with other state evaluation methods, the method has fault tolerance and tolerance capabilities and strong self-adaption capability; the method has the characteristics of high recognition speed, stable learning algorithm, easy training and the like. Actual test data shows that the average prediction accuracy of the model established by the method can reach 89.12 percent, and is higher than that of a decision tree model and a support vector machine model, so that the reliability of the model is stronger. The method has the advantages that the possible defects of the lightning arrester are inferred and predicted, more effective decision basis can be provided for the site, and the method has certain engineering application value.
Claims (2)
1. A lightning arrester insulation state assessment method based on random forests is characterized by comprising the following steps:
s1, acquiring samples to be tested of the lightning arrester in different insulation state types, wherein the insulation state types of the lightning arrester can be roughly divided into four types, namely internal damping, insulation aging, external insulation dirt and other types according to historical fault data samples and field actual operation experience;
s2, the characteristic data of the lightning arrester mainly comprise live detection information, online monitoring information, material detection information and environment information;
and S3, carrying out data normalization processing on the data set obtained in the S2 by adopting a z-score normalization method, wherein the calculation formula is as follows:
the average value of new data obtained after the z-score normalization treatment is 0, and the standard deviation is 1;
s4, segmenting the normalized data, wherein a part of data is used as a test set, and a part of data is used as a training set;
s5, constructing an arrester insulation state evaluation model on the training set by using a random forest algorithm, and simultaneously obtaining two key parameters of the random forest by using a Grid Search (Grid Search) and cross validation mode: the optimal values of the number of features per tree (Mfeature) and the number of trees (Ntree) enable the stochastic tree model to better fit the training samples;
the method of cross validation in S5 is:
randomly dividing a training sample into m mutually disjoint subsets S1, S2, S3, \8230;. Sm, wherein the size of each subset is approximately equal, m-1 parts are selected as a training set each time, and the rest 1 part is used as a test set for training and testing; training and testing are carried out for m times, each sub-sample is tested once, different test effects for m times are obtained, and the average result for m times is used as the final model effect; the grid searching method in the S5 comprises the following steps:
the random forest method has two key parameters, the effectiveness of the model is greatly influenced by the value of the two key parameters, namely the characteristic quantity (Mfeature) of each tree and the number (Ntree) of the trees;
setting the value intervals of the characteristic quantity Mfeature and the number Ntree of the trees, setting corresponding step length, traversing all parameter combinations, combining cross validation, calculating the size of the area (ROC _ auc) under the evaluation index ROC curve of each combination, and determining the optimal characteristic quantity Mfeature and the number Ntree of the trees, thereby obtaining the optimal arrester insulation state classification prediction model;
s6, based on the test set data in the S5, evaluating the model obtained by training by using indexes such as overall accuracy, recall ratio, precision ratio and the like;
and S7, if the evaluation index of S6 meets the requirement, inputting the sample to be tested in S1 into an arrester insulation state evaluation model, and obtaining the insulation state type of the sample to be tested.
2. A random forest based lightning arrester insulation state assessment method according to claim 1, characterised in that the random forest algorithm comprises the steps of:
s51, selecting a sample from the original data set as a training set by a sampling method (bootstrap) with feedback in the training data set;
s52, generating a decision tree by using the training set obtained by sampling, wherein when each node of the tree is generated: the k features are selected randomly and non-repetitively, typically
The M is the number of characteristic variables;
the k characteristics are used for dividing the sample set respectively to find the optimal division characteristics, and the optimal division characteristics can be distinguished by a Gini coefficient, a gain rate and an information gain index;
s53, repeating the steps S51-S52 for K times, wherein K is the number of decision trees in the random forest;
and S54, establishing a random forest classification model, and obtaining a final prediction result by adopting a majority voting method because the state evaluation is a classification problem, namely voting the prediction results of all decision trees in the random forest model, wherein the highest obtained category is the final output category.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211268533.9A CN115561596A (en) | 2022-10-17 | 2022-10-17 | Lightning arrester insulation state assessment method based on random forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211268533.9A CN115561596A (en) | 2022-10-17 | 2022-10-17 | Lightning arrester insulation state assessment method based on random forest |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115561596A true CN115561596A (en) | 2023-01-03 |
Family
ID=84746756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211268533.9A Pending CN115561596A (en) | 2022-10-17 | 2022-10-17 | Lightning arrester insulation state assessment method based on random forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115561596A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910668A (en) * | 2023-09-11 | 2023-10-20 | 国网浙江省电力有限公司余姚市供电公司 | Lightning arrester fault early warning method, device, equipment and storage medium |
-
2022
- 2022-10-17 CN CN202211268533.9A patent/CN115561596A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116910668A (en) * | 2023-09-11 | 2023-10-20 | 国网浙江省电力有限公司余姚市供电公司 | Lightning arrester fault early warning method, device, equipment and storage medium |
| CN116910668B (en) * | 2023-09-11 | 2024-04-02 | 国网浙江省电力有限公司余姚市供电公司 | Lightning arrester fault early warning method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106908674B (en) | Transformer state evaluation method based on multi-state quantity prediction | |
| CN103366123B (en) | Software hazard appraisal procedure based on defect analysis | |
| CN109217296B (en) | Power grid key node identification method based on weighted power flow entropy and voltage stability | |
| CN111210075B (en) | Lightning transmission line fault probability analysis method based on combined classifier | |
| CN110208647B (en) | Power distribution network single-phase earth fault section positioning method based on random matrix | |
| CN115048985B (en) | Electrical equipment fault discrimination method | |
| CN115561596A (en) | Lightning arrester insulation state assessment method based on random forest | |
| CN111177655B (en) | Data processing method and device and electronic equipment | |
| CN116739829B (en) | Big data-based power data analysis method, system and medium | |
| CN116882790B (en) | Carbon emission equipment management method and system for mine ecological restoration area | |
| CN116366374B (en) | Security assessment method, system and medium for power grid network management based on big data | |
| CN115936262B (en) | Yield prediction method, system and medium based on big data environment interference | |
| CN116737510A (en) | Data analysis-based intelligent keyboard monitoring method and system | |
| CN110930057A (en) | Quantitative evaluation method for reliability of distribution transformer test result based on LOF algorithm | |
| CN112633331B (en) | Thermal power plant auxiliary machine health state assessment method based on self-encoder | |
| CN117368651B (en) | Comprehensive analysis system and method for faults of power distribution network | |
| CN112597435B (en) | Thermal power equipment quality data processing method and device based on equipment supervision | |
| CN113591402A (en) | Digital power transformer health state fuzzy comprehensive evaluation method | |
| CN110807014B (en) | Cross validation based station data anomaly discrimination method and device | |
| CN115729761B (en) | Hard disk fault prediction method, system, equipment and medium | |
| CN112016248A (en) | Elman neural network based SCR denitration system bad data identification method | |
| CN111988814A (en) | Method for evaluating link quality by adopting improved variational self-encoder | |
| CN114580973B (en) | Quality monitoring method and system in transformer production process | |
| CN116151799A (en) | BP neural network-based distribution line multi-working-condition fault rate rapid assessment method | |
| CN116662900A (en) | Partial discharge degree identification method and system for low-frequency voltage |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20231024 Address after: North Office Building, No. 12, Nanxiaoqiang, Xinghualing District, Taiyuan City, Shanxi Province, 030000 Applicant after: MATERIAL BRANCH OF STATE GRID SHANXI ELECTRIC POWER Co. Applicant after: STATE GRID ELECTRIC POWER Research Institute OF SEPC Address before: North Office Building, No. 12, Nanxiaoqiang, Xinghualing District, Taiyuan City, Shanxi Province, 030001 Applicant before: MATERIAL BRANCH OF STATE GRID SHANXI ELECTRIC POWER Co. |
|
| TA01 | Transfer of patent application right |