CN112308146A - Distribution transformer fault identification method based on operation characteristics - Google Patents
Distribution transformer fault identification method based on operation characteristics Download PDFInfo
- Publication number
- CN112308146A CN112308146A CN202011200609.5A CN202011200609A CN112308146A CN 112308146 A CN112308146 A CN 112308146A CN 202011200609 A CN202011200609 A CN 202011200609A CN 112308146 A CN112308146 A CN 112308146A
- Authority
- CN
- China
- Prior art keywords
- distribution transformer
- distribution
- fault
- data
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention relates to a distribution transformer fault identification method based on operation characteristics. The method comprises the following steps: step S1, on the basis of the distribution transformer body fault power failure record, combining the operation characteristic data before the distribution transformer fault, including distribution transformer archive parameters, operation data and environment data, screening the power failure record conforming to the distribution transformer body fault, positioning the power failure record to a specific distribution transformer equipment ID, and constructing a distribution transformer data set; s2, extracting important characteristic variables influencing distribution transformer faults by using a random forest algorithm based on a distribution transformer data set, and then constructing a distribution transformer fault early warning model by using a machine learning algorithm; and S3, outputting a distribution transformer operation risk grade based on the distribution transformer fault early warning model, and positioning a fault part.
Description
Technical Field
The invention relates to a distribution transformer fault identification method based on operation characteristics.
Background
Based on the distribution transformer equipment health state perception and historical fault information data analysis, the occurrence reason of the abnormal state of the distribution transformer equipment and the potential operation rule of the distribution transformer in the fault period are mined, the early signs of the fault are identified, the fault position, the fault degree and the development trend are researched and judged, and the optimal operation maintenance and overhaul time can be determined.
Disclosure of Invention
The invention aims to provide a distribution transformer fault identification method based on operation characteristics, which can identify early signs of faults, study and judge fault positions, fault degrees and development trends and determine the optimal operation, maintenance and overhaul time.
In order to achieve the purpose, the technical scheme of the invention is as follows: a distribution transformer fault identification method based on operation characteristics comprises the following steps:
step S1, on the basis of the distribution transformer body fault power failure record, combining the operation characteristic data before the distribution transformer fault, including distribution transformer archive parameters, operation data and environment data, screening the power failure record conforming to the distribution transformer body fault, positioning the power failure record to a specific distribution transformer equipment ID, and constructing a distribution transformer data set;
s2, extracting important characteristic variables influencing distribution transformer faults by using a random forest algorithm based on a distribution transformer data set, and then constructing a distribution transformer fault early warning model by using a machine learning algorithm;
and S3, outputting a distribution transformer operation risk grade based on the distribution transformer fault early warning model, and positioning a fault part.
In an embodiment of the present invention, the specific manner of extracting the important characteristic variables affecting the distribution transformer fault by using the random forest algorithm in step S2 is as follows: firstly, randomly extracting half of data from a distribution transformation data set as a classification regression tree, and taking the rest half of the data as data outside a bag; secondly, randomly extracting half of feature variables at each node of each number, calculating the information content contained in each feature, and selecting the maximum value of the information content as the node splitting of the first feature; then, arranging the data in descending order according to the information quantity, and stopping splitting when the error value is minimum; and finally, selecting a characteristic variable set with the largest overall information quantity and the smallest error as a core characteristic variable, namely an important characteristic variable influencing the distribution transformer fault.
In an embodiment of the present invention, the machine learning algorithm adopted in step S2 is Adaboost algorithm.
In an embodiment of the invention, a mode for constructing a distribution transformer fault early warning model by using an Adaboost algorithm is as follows:
defining a distribution data set T { (x)1,y1),(x2,y2)...(xN,yN) Therein examplesBut example spaceyiBelongs to the label set { -1, +1 };
(1) initializing weight distribution of data in a distribution transformation data set, namely training data, wherein each training data sample is endowed with the same weight at the beginning: 1/N;
in the formula, D1Represents the weight, w, of each sample of the first iteration1iRepresenting the weight of the ith sample in the 1 st iteration, wherein N is the total number of samples;
(2) a plurality of iterations are performed, with M being 1, 2.
A. Using a weight distribution DmLearning the distribution transformation data set to obtain a basic classifier:
Gm(x):χ→{-1,+1}
this equation represents the basic classifier G at the mth iterationm(x) Sample x is classified as either-1 or 1, Gm(x) Is a two-classifier;
B. calculation of Gm(x) Classification error rate on the distribution data set:
C. calculation of Gm(x) Coefficient of (a)mRepresents Gm(x) Importance in the final classifier):
from the above formula, emWhen < 1/2, am0, and amWith emIs increased, meaning that the smaller the classification error rate, the more the basic classifier plays a role in the final classifier;
D. updating weight distribution of the distribution transformation data set so as to enable the distribution transformation data set to be classified by a basic classifier Gm(x) The weight of the misclassified samples is increased, while the weight of the correctly classified samples is decreased;
Dm+1=(wm+1,1,wm+1,2…wm+1,i…,wm+1,N),
wherein Z ismIs a normalization factor such that Dm+1Becomes a probability distribution
(3) Combining the weak classifiers:
and further, obtaining a final classifier, namely a distribution fault early warning model:
compared with the prior art, the invention has the following beneficial effects: the method can identify the early signs of the fault, study and judge the fault position, the fault degree and the development trend, and determine the optimal operation and maintenance time.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a distribution transformer fault operation characteristic data processing process.
Fig. 3 is a process for constructing a distribution transformer equipment fault operation characteristic model.
Detailed Description
The technical scheme of the invention is specifically explained below with reference to the accompanying drawings.
As shown in fig. 1, the present invention provides a distribution transformer fault identification method based on operation characteristics, including the following steps:
step S1, on the basis of the distribution transformer body fault power failure record, combining the operation characteristic data before the distribution transformer fault, including distribution transformer archive parameters, operation data and environment data, screening the power failure record conforming to the distribution transformer body fault, positioning the power failure record to a specific distribution transformer equipment ID, and constructing a distribution transformer data set;
s2, extracting important characteristic variables influencing distribution transformer faults by using a random forest algorithm based on a distribution transformer data set, and then constructing a distribution transformer fault early warning model by using a machine learning algorithm;
and S3, outputting a distribution transformer operation risk grade based on the distribution transformer fault early warning model, and positioning a fault part.
The following is a specific implementation of the present invention.
The invention relates to a distribution transformer fault identification method based on operation characteristics, which is realized as follows:
as shown in fig. 2, firstly, correlation analysis is performed by combining pre-distribution-fault operation characteristic data such as distribution transformer archive parameters, operation data, environmental data and the like on the basis of the distribution transformer body fault power failure records. In this example, since the fault information pointing to the specific distribution transformer cannot be directly acquired, it is necessary to screen the power failure records conforming to the fault of the distribution transformer body by searching the fault summary in the fault records, locate the specific equipment ID, and finally acquire about 290 public distribution transformer body fault power failure records conforming to the condition.
As shown in fig. 3, then, based on multi-group algorithm exploration and comparison, the interpretability of the random forest and the Adaboost algorithm is high, the implementation difficulty is low, and the prediction accuracy is high. Important characteristic variables influencing the distribution transformer fault are extracted by using a random forest algorithm, a distribution transformer fault early warning model is constructed by using machine learning algorithms such as adaboost and the like, the distribution transformer operation risk level is output, the fault position is positioned, and the equipment safety is guaranteed. The method comprises the following specific steps:
(1) random forest feature screening
The specific steps of the random forest characteristic screening are as follows: firstly, randomly extracting half of data from a training set as a classification regression tree, and taking the remaining half of the data as data outside a bag; secondly, randomly extracting half of feature variables at each node of each number, calculating the information content contained in each feature, and selecting the maximum value of the information content as the node splitting of the first feature; then, the division is stopped when the error value is minimum. And finally, selecting a characteristic variable set with the largest overall information quantity and the smallest error as a core characteristic variable.
(2)Adaboost
The algorithm flow of Adaboost is as follows:
defining a distribution data set T { (x)1,y1),(x2,y2)...(xN,yN) Therein examplesBut example spaceyiThe goal of Adaboost, belonging to the label set { -1, +1}, is to learn a series of weak classifiers, or basic classifiers, from the training data and then combine these weak classifiers into one strong classifier.
Step 1, firstly, initializing weight distribution of training data. Each training sample is initially given the same weight: 1/N.
And 2, performing multiple iterations, and using M as 1, 2, wherein M represents the first iteration round
a. Using a weight distribution DmLearning the training data set to obtain a basic classifier (selecting a threshold value with the lowest error rate to design the basic classifier):
Gm(x):χ→{-1,+1}
b. calculation of Gm(x) Classification error rate on training data set
From the above formula, Gm(x) Error rate on training data set emIs exactly covered by Gm(x) The sum of the weights of the misclassified samples.
c. Calculation of Gm(x) Coefficient of (a)mRepresents Gm(x) Degree of importance in the final classifier (purpose: get the weight the basic classifier takes in the final classifier):
from the above formula, emWhen < 1/2, am0, and amWith emIs increased, means that the basic classifier with a smaller classification error rate has a higher role in the final classifier.
d. The weight distribution of the training data set is updated (in order to obtain a new weight distribution of the samples) for the next iteration
Dm+1=(wm+1,1,wm+1,2…wm+1,i…,wm+1,N),
So as to be classified by the basic classifier Gm(x) The weight of misclassified samples increases and the weight of correctly classified samples decreases. In this manner, the AdaBoost method can "focus" or "focus on" those samples that are less readily separable.
Wherein Z ismIs a normalization factor such that Dm+1Becomes a probability distribution:
step 3, combining each weak classifier
The final classifier is thus obtained as follows:
the above are preferred embodiments of the present invention, and all changes made according to the technical scheme of the present invention that produce functional effects do not exceed the scope of the technical scheme of the present invention belong to the protection scope of the present invention.
Claims (4)
1. A distribution transformer fault identification method based on operation characteristics is characterized by comprising the following steps:
step S1, on the basis of the distribution transformer body fault power failure record, combining the operation characteristic data before the distribution transformer fault, including distribution transformer archive parameters, operation data and environment data, screening the power failure record conforming to the distribution transformer body fault, positioning the power failure record to a specific distribution transformer equipment ID, and constructing a distribution transformer data set;
s2, extracting important characteristic variables influencing distribution transformer faults by using a random forest algorithm based on a distribution transformer data set, and then constructing a distribution transformer fault early warning model by using a machine learning algorithm;
and S3, outputting a distribution transformer operation risk grade based on the distribution transformer fault early warning model, and positioning a fault part.
2. The distribution transformer fault identification method based on the operation characteristics as claimed in claim 1, wherein the specific way of extracting the important characteristic variables affecting the distribution transformer fault by using the random forest algorithm in step S2 is as follows: firstly, randomly extracting half of data from a distribution transformation data set as a classification regression tree, and taking the rest half of the data as data outside a bag; secondly, randomly extracting half of feature variables at each node of each number, calculating the information content contained in each feature, and selecting the maximum value of the information content as the node splitting of the first feature; then, arranging the data in descending order according to the information quantity, and stopping splitting when the error value is minimum; and finally, selecting a characteristic variable set with the largest overall information quantity and the smallest error as a core characteristic variable, namely an important characteristic variable influencing the distribution transformer fault.
3. The distribution transformer fault identification method based on the operation characteristics as claimed in claim 1, wherein the machine learning algorithm adopted in step S2 is Adaboost algorithm.
4. The distribution transformer fault identification method based on the operation characteristics as claimed in claim 3, wherein the mode of constructing the distribution transformer fault early warning model by using Adaboost algorithm is as follows:
defining a distribution data set T { (x)1,y1),(x2,y2)...(xN,yN) Therein examplesBut example spaceyiBelongs to the label set { -1, +1 };
(1) initializing weight distribution of data in a distribution transformation data set, namely training data, wherein each training data sample is endowed with the same weight at the beginning: 1/N;
in the formula, D1Represents the weight, w, of each sample of the first iteration1iRepresenting the weight of the ith sample in the 1 st iteration, wherein N is the total number of samples;
(2) multiple iterations are performed, where M is 1, 2, …, and M denotes the number of iterations:
A. using a weight distribution DmLearning the distribution transformation data set to obtain a basic classifier:
Gm(x):χ→{-1,+1}
this equation represents the basic classifier G at the mth iterationm(x) Sample x is classified as either-1 or 1, Gm(x) Is a two-classifier;
B. calculation of Gm(x) Classification error rate on the distribution data set:
C. calculation of Gm(x) Coefficient of (a)mRepresents Gm(x) InImportance in the final classifier:
from the above formula, emWhen < 1/2, am0, and amWith emIs increased, meaning that the smaller the classification error rate, the more the basic classifier plays a role in the final classifier;
D. updating weight distribution of the distribution transformation data set so as to enable the distribution transformation data set to be classified by a basic classifier Gm(x) The weight of the misclassified samples is increased, while the weight of the correctly classified samples is decreased;
Dm+1=(wm+1,1,wm+1,2…wm+1,i…,wm+1,N),
wherein Z ismIs a normalization factor such that Dm+1Becomes a probability distribution
(3) Combining the weak classifiers:
and further, obtaining a final classifier, namely a distribution fault early warning model:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011200609.5A CN112308146A (en) | 2020-11-02 | 2020-11-02 | Distribution transformer fault identification method based on operation characteristics |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011200609.5A CN112308146A (en) | 2020-11-02 | 2020-11-02 | Distribution transformer fault identification method based on operation characteristics |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112308146A true CN112308146A (en) | 2021-02-02 |
Family
ID=74333322
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011200609.5A Pending CN112308146A (en) | 2020-11-02 | 2020-11-02 | Distribution transformer fault identification method based on operation characteristics |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112308146A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107179503A (en) * | 2017-04-21 | 2017-09-19 | 美林数据技术股份有限公司 | The method of Wind turbines intelligent fault diagnosis early warning based on random forest |
CN107730059A (en) * | 2017-11-29 | 2018-02-23 | 成都思晗科技股份有限公司 | The method of transformer station's electricity trend prediction analysis based on machine learning |
CN110108992A (en) * | 2019-05-24 | 2019-08-09 | 国网湖南省电力有限公司 | Based on cable partial discharge fault recognition method, system and the medium for improving random forests algorithm |
AU2020100709A4 (en) * | 2020-05-05 | 2020-06-11 | Bao, Yuhang Mr | A method of prediction model based on random forest algorithm |
CN111860658A (en) * | 2020-07-24 | 2020-10-30 | 华北电力大学(保定) | Transformer fault diagnosis method based on cost sensitivity and integrated learning |
-
2020
- 2020-11-02 CN CN202011200609.5A patent/CN112308146A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107179503A (en) * | 2017-04-21 | 2017-09-19 | 美林数据技术股份有限公司 | The method of Wind turbines intelligent fault diagnosis early warning based on random forest |
CN107730059A (en) * | 2017-11-29 | 2018-02-23 | 成都思晗科技股份有限公司 | The method of transformer station's electricity trend prediction analysis based on machine learning |
CN110108992A (en) * | 2019-05-24 | 2019-08-09 | 国网湖南省电力有限公司 | Based on cable partial discharge fault recognition method, system and the medium for improving random forests algorithm |
AU2020100709A4 (en) * | 2020-05-05 | 2020-06-11 | Bao, Yuhang Mr | A method of prediction model based on random forest algorithm |
CN111860658A (en) * | 2020-07-24 | 2020-10-30 | 华北电力大学(保定) | Transformer fault diagnosis method based on cost sensitivity and integrated learning |
Non-Patent Citations (1)
Title |
---|
程淼海;楼俏;王琼;王国军;胡殿刚;李韶瑜;: "基于随机森林算法的配网抢修故障量预测方法", 计算机系统应用, no. 09 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111274395B (en) | Power grid monitoring alarm event identification method based on convolution and long-short term memory network | |
CN111914873A (en) | Two-stage cloud server unsupervised anomaly prediction method | |
US11840998B2 (en) | Hydraulic turbine cavitation acoustic signal identification method based on big data machine learning | |
CN106570513A (en) | Fault diagnosis method and apparatus for big data network system | |
CN111143838B (en) | Database user abnormal behavior detection method | |
CN112199670B (en) | Log monitoring method for improving IFOREST (entry face detection sequence) to conduct abnormity detection based on deep learning | |
CN111027629A (en) | Power distribution network fault outage rate prediction method and system based on improved random forest | |
CN113438114B (en) | Method, device, equipment and storage medium for monitoring running state of Internet system | |
CN108647772B (en) | Method for removing gross errors of slope monitoring data | |
CN115021679A (en) | Photovoltaic equipment fault detection method based on multi-dimensional outlier detection | |
CN114609994A (en) | Fault diagnosis method and device based on multi-granularity regularization rebalance incremental learning | |
CN110766215B (en) | Wind power climbing event prediction method based on feature adaptive selection and WDNN | |
CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
CN116400168A (en) | Power grid fault diagnosis method and system based on depth feature clustering | |
CN111107082A (en) | Immune intrusion detection method based on deep belief network | |
CN112308146A (en) | Distribution transformer fault identification method based on operation characteristics | |
CN115965135A (en) | New energy prediction error modeling method and system based on naive Bayes classification | |
CN112465054B (en) | FCN-based multivariate time series data classification method | |
CN112541554B (en) | Multi-mode process monitoring method and system based on time constraint and nuclear sparse representation | |
CN117592789B (en) | Power grid environment fire risk assessment method and equipment based on time sequence analysis | |
US20230012643A1 (en) | Method for predicting c-axis length of lithium compound crystal structure, method for building learning model, and system for predicting crystal structure having maximum c-axis length | |
CN115288994B (en) | Improved DCGAN-based compressor abnormal state detection method | |
CN117435904B (en) | Single feature ordering and composite feature extraction method | |
Wu et al. | A comprehensive modeling method of continuous and discrete variables for personal credit forecasting | |
Fu et al. | Anomaly Detection Method Based on Gaussian Mixture Model and Orthogonal Generative Adversarial Network |
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 |