CN110245617B - Artificial intelligence analysis method based on transient recording waveform - Google Patents
Artificial intelligence analysis method based on transient recording waveform Download PDFInfo
- Publication number
- CN110245617B CN110245617B CN201910522198.2A CN201910522198A CN110245617B CN 110245617 B CN110245617 B CN 110245617B CN 201910522198 A CN201910522198 A CN 201910522198A CN 110245617 B CN110245617 B CN 110245617B
- Authority
- CN
- China
- Prior art keywords
- subset
- data
- waveform
- waveform data
- artificial intelligence
- 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.)
- Active
Links
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
- G01R31/08—Locating faults in cables, transmission lines, or networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention provides an artificial intelligence analysis method based on a transient recording waveform, which comprises the following steps: step S1, transient waveform with disturbance is collected from the power grid environment, and after the transient wave recording collector is triggered, the waveform data of a plurality of periods before the triggering moment and the waveform data of a plurality of periods after the triggering moment are recorded; the collected waveform data samples comprise six variables of A, B, C three-phase voltage and A, B, C three-phase current; step S2, sampling the acquired waveform data, and sampling a plurality of data points in each period; and step S3, inputting the sampled waveform data into a trained artificial intelligence algorithm model, wherein the artificial intelligence algorithm model gives the waveform type. The invention improves the waveform recognition efficiency of the power grid fault, saves manpower and reduces the cost of operation and maintenance.
Description
Technical Field
The invention relates to the field of transient recording waveforms of a power grid, in particular to an artificial intelligence analysis method based on transient recording waveforms.
Background
At present, common methods and practical detection devices are used for detecting certain steady states such as voltage deviation, frequency deviation, steady harmonic distortion, voltage fluctuation and flicker in a power grid. However, automated detection and analysis of power system waveform transients such as voltage sag, voltage swell, periodic notches, etc. have consistently lacked an effective solution. The current processing means is to manually compare the waveform obtained by recording with the corresponding normal waveform point to judge the type of transient recording.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides an artificial intelligence analysis method based on a transient recording waveform, and improves the waveform identification efficiency of power grid faults. The technical scheme adopted by the invention is as follows:
an artificial intelligence analysis method based on transient recording waveforms comprises the following steps:
step S1, transient waveform with disturbance is collected from the power grid environment, and after the transient wave recording collector is triggered, the waveform data of a plurality of periods before the triggering moment and the waveform data of a plurality of periods after the triggering moment are recorded;
the collected waveform data samples comprise six variables of A, B, C three-phase voltage and A, B, C three-phase current;
step S2, sampling the acquired waveform data, and sampling a plurality of data points in each period;
and step S3, inputting the sampled waveform data into a trained artificial intelligence algorithm model, wherein the artificial intelligence algorithm model gives the waveform type.
Further, the training process of the artificial intelligence algorithm model is as follows:
step 1, collecting waveform data samples in advance, wherein the waveform data samples comprise three types of waveform data, namely normal waveform data, short-circuit waveform data and grounding waveform data, and each type of waveform data sample comprises six variables, namely A, B, C three-phase voltage and A, B, C three-phase current;
step 2, calculating residual error data of each variable, and subtracting data points in subsequent periods of the variable from corresponding position data points in a first period in a sample by taking a data point of the first period of the variable as a reference to obtain the residual error data of the variable;
step 3, calculating a characteristic value: calculating the skewness, kurtosis, range, mean value and standard deviation of six variables of A, B, C three-phase voltage and A, B, C three-phase current of a waveform data sample, and respectively counting the number of data points of which the value obtained by subtracting the mean value from residual data in each variable is more than 2 times of the standard deviation and the number of data points of which the value is more than 3 times of the standard deviation; the calculated skewness, kurtosis, range and residual data minus the mean value are used as the input characteristics of the artificial intelligence algorithm model, wherein the data points are more than 2 times of the standard deviation and more than 3 times of the standard deviation;
step 4, constructing a random forest consisting of a plurality of decision trees, and distinguishing input features by taking the information entropy as a standard for feature selection;
the construction process of the decision tree comprises the following steps:
step 401, firstly, calculating the initial information entropy of the current sample data set, and calculating the initial information entropy before the subset is not divided according to the original sample data set D;
step 402, then calculating the information entropy of 30 characteristics of 6 variables of each sample; the implementation process is to use a binary tree to divide the current sample data set into two subsets DleftAnd DrightThe characteristic information entropy formula:
the method comprises the following steps that N represents the number of samples in a current sample data set, the number of samples in D is represented for an original sample data set D, and the number of samples in the subset is represented if the current sample data set is a divided subset; n is a radical ofleftRepresenting a subset D of the current sample data setleftNumber of middle samples, NrightAnother subset D representing the separation of the current sample data setrightThe number of the middle samples; e () represents the entropy of the calculation information;
dividing two subsets DleftAnd DrightThe method comprises arranging an input feature in the order from small to large, sequentially taking the feature value of the feature, and classifying the feature value into DleftIn the subset, greater than the eigenvalues are assigned to DrightSubset, calculating all corresponding I of all possible subset division conditions of the feature, and taking the minimum I as the information entropy of the feature;
calculating the information entropy of all the characteristics;
step 403, calculate In-Einit，InExpressing the information entropy of the nth feature, taking In-EinitDividing the current sample data set into two subsets by taking the characteristic with the maximum value as a segmentation point;
step 404, recursively calling steps 401 to 403 in the subset part until the entropy of the subset information is 0 or In-EinitWhen the number of the subsets is smaller than the threshold value, stopping continuously dividing the subsets, and finishing the model training;
in the subset at the lowest layer, the type with the largest number of sample types represents the type of the subset;
the artificial intelligence algorithm model returns the type of the subset corresponding to the input features according to the input features.
The invention has the advantages that: the invention improves the waveform recognition efficiency of the power grid fault, saves manpower, reduces the cost of operation and maintenance and also reduces the misjudgment probability.
Drawings
Fig. 1 is a schematic diagram of the training process of the present invention.
Detailed Description
The invention is further illustrated by the following specific figures and examples.
An artificial intelligence analysis method based on transient recording waveforms comprises the following steps:
step S1, transient waveform with disturbance is collected from the power grid environment, and after the transient wave recording collector is triggered, waveform data of 4 periods before the triggering moment and waveform data of 8 periods after the triggering moment are recorded;
the collected waveform data samples comprise six variables of A, B, C three-phase voltage and A, B, C three-phase current;
step S2, sampling the acquired waveform data, for example, sampling 80 data points per cycle;
then one variable samples 960 data points for 12 cycles;
step S3, inputting the sampled waveform data into a trained artificial intelligence algorithm model, wherein the artificial intelligence algorithm model gives a waveform type, such as normal, short circuit or grounding;
the training process of the artificial intelligence algorithm model is as follows:
step 1, collecting waveform data samples in advance, wherein the waveform data samples comprise three types of waveform data, namely normal waveform data, short-circuit waveform data and grounding waveform data, and each type of waveform data sample comprises six variables, namely A, B, C three-phase voltage and A, B, C three-phase current;
step 2, calculating residual error data of each variable, and subtracting data points in subsequent periods of the variable from corresponding position data points in a first period in a sample by taking a data point of the first period of the variable as a reference to obtain the residual error data of the variable;
taking the phase voltage A as an example, taking 80 data points in the first period of the phase voltage A as a reference, and subtracting the corresponding position data point in the first period from 80 data points in each period from the second period to the 12 th period to obtain residual data of the phase voltage A;
step 3, calculating a characteristic value: calculating the skewness, kurtosis, pole difference, mean value and standard deviation of six variables of A, B, C three-phase voltage and A, B, C three-phase current of a waveform data sample, and respectively counting the number of data points of which the value obtained by subtracting the mean value from residual data in each variable is more than 2 times of the standard deviation and the number of data points of which the value is more than 3 times of the standard deviation; the calculated skewness, kurtosis, range and residual data minus the mean value are used as the input characteristics of the artificial intelligence algorithm model, wherein the data points are more than 2 times of the standard deviation and more than 3 times of the standard deviation;
step 4, constructing a random forest consisting of a plurality of decision trees, and distinguishing input features by taking the information entropy as a standard for feature selection;
the information entropy is a most commonly used index for measuring the purity of a sample set;
suppose that the ratio of the kth sample in the current sample set D is pk(k is 1, 2, …, n), the entropy of information of D is defined as
The smaller the value of E (D), the higher the purity of D;
the construction process of the decision tree comprises the following steps:
step 401, firstly, calculating the initial information entropy of the current sample data set, and calculating the initial information entropy before the subset is not divided according to the original sample data set D; the waveform is divided into three types, namely normal p1, short circuit p2 and grounding p 3;
initial information entropy Einit＝-p1log2p1-p2log2p2-p3log2p3；
Step 402, then calculating the information entropy of 30 characteristics of 6 variables of each sample; fruit of Chinese wolfberryThe present process uses a binary tree to divide the current sample data set into two subsets DleftAnd DrightThe characteristic information entropy formula:
the N represents the number of samples in the current sample data set, the number of samples in the original sample data set D is represented, and if the current sample data set is a divided subset, the number of samples in the subset is represented; n is a radical ofleftRepresenting a subset D of the current sample data setleftNumber of middle samples, NrightAnother subset D representing the separation of the current sample data setrightThe number of the middle samples; e () represents the computation information entropy;
dividing two subsets DleftAnd DrightThe method comprises arranging an input feature in the order from small to large, sequentially taking the feature value of the feature, and classifying the feature value into DleftIn the subset, greater than this eigenvalue falls to DrightSubset, calculating all corresponding I of all possible subset division conditions of the feature, and taking the minimum I as the information entropy of the feature;
calculating the information entropy of all the characteristics;
step 403, calculate In-Einit，InExpressing the information entropy of the nth feature, taking In-EinitDividing the current sample data set into two subsets by taking the characteristic with the maximum value as a segmentation point;
step 404, recursively calling steps 401 to 403 in the subset part until the entropy of the subset information is 0 or In-EinitWhen the number of the subsets is smaller than the threshold value, stopping continuously dividing the subsets, and finishing the model training;
in the lowest subset, the type with the largest number of sample types represents the type of the subset.
The artificial intelligence algorithm model returns the type of subset corresponding to the input features, such as normal, short, or ground, based on the input features.
One example is shown in figure 1 of the drawings,
1) firstly, calculating initial information entropy E corresponding to original sample data set DinitAnd information entropy I of each featuren，In-EinitThe maximum difference value is the characteristic of 'extreme difference of A phase voltage residual errors', and the characteristic is selected as a dividing point;
2) dividing the characteristic that the value of the 'A phase voltage residual extreme difference' is greater than 10 into a subset D1, and dividing the characteristic that the value of the 'A phase voltage residual extreme difference' is less than or equal to 10 into a subset D2;
3) calculating the information entropy E of D1initAnd entropy I of each feature information in D1n，In-EinitThe maximum difference value is the characteristic of 'extreme difference of B-phase voltage residual error', and the characteristic is selected as a dividing point;
4) dividing the characteristic that the value of the 'B phase voltage residual extreme difference' is greater than 8.5 into a subset D3, and dividing the characteristic that the value of the characteristic is less than or equal to 8.5 into a subset D4;
5) the division of the subset D2 branch is similar to that of D1;
6) and finally, four subsets D3, D4, D5 and D6 which meet the information entropy condition are obtained, and in each subset, the type with the largest number of sample types represents the type of the subset.
Finally, it should be noted that the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention has been described in detail with reference to examples, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, which should be covered by the claims of the present invention.
Claims (2)
1. An artificial intelligence analysis method based on transient recording waveforms is characterized by comprising the following steps:
step S1, transient waveform with disturbance is collected from the power grid environment, and after the transient wave recording collector is triggered, waveform data of a plurality of periods before the triggering moment and waveform data of a plurality of periods after the triggering moment are recorded;
the collected waveform data samples comprise six variables of A, B, C three-phase voltage and A, B, C three-phase current;
step S2, sampling the acquired waveform data, and sampling a plurality of data points in each period;
step S3, inputting the sampled waveform data into a trained artificial intelligence algorithm model, wherein the artificial intelligence algorithm model gives a waveform type;
the training process of the artificial intelligence algorithm model is as follows:
step 1, collecting waveform data samples in advance, wherein the waveform data samples comprise three types of waveform data, namely normal waveform data, short-circuit waveform data and grounding waveform data, and each type of waveform data sample comprises six variables, namely A, B, C three-phase voltage and A, B, C three-phase current;
step 2, calculating residual error data of each variable, and subtracting data points in subsequent periods of the variable from corresponding position data points in a first period in a sample by taking a data point of the first period of the variable as a reference to obtain the residual error data of the variable;
step 3, calculating a characteristic value: calculating the skewness, kurtosis, range, mean value and standard deviation of six variables of A, B, C three-phase voltage and A, B, C three-phase current of a waveform data sample, and respectively counting the number of data points of which the value obtained by subtracting the mean value from residual data in each variable is more than 2 times of the standard deviation and the number of data points of which the value is more than 3 times of the standard deviation; the calculated skewness, kurtosis, range and residual data minus the mean value are used as the input characteristics of the artificial intelligence algorithm model, wherein the data points are more than 2 times of the standard deviation and more than 3 times of the standard deviation;
step 4, constructing a random forest consisting of a plurality of decision trees, and distinguishing input features by taking the information entropy as a standard for feature selection;
the construction process of the decision tree comprises the following steps:
step 401, firstly, calculating the initial information entropy of the current sample data set, and calculating the initial information entropy before the subset is not divided according to the original sample data set D;
step 402, then calculating the information entropy of 30 characteristics of 6 variables of each sample; the implementation process is to use a binary treeDividing the current sample data set into two subsets DleftAnd DrightThe characteristic information entropy formula:
the N represents the number of samples in the current sample data set, the number of samples in the original sample data set D is represented, and if the current sample data set is a divided subset, the number of samples in the subset is represented; n is a radical ofleftA subset D representing the current sample data setleftNumber of middle samples, NrightAnother subset D representing the separation of the current sample data setrightThe number of the middle samples; e () represents the computation information entropy;
dividing two subsets DleftAnd DrightThe method comprises arranging an input feature in the order from small to large, sequentially taking the feature value of the feature, and classifying the feature value into DleftIn the subset, greater than the eigenvalues are assigned to DrightSubset, calculating all corresponding I of all possible subset division conditions of the feature, and taking the minimum I as the information entropy of the feature;
calculating the information entropy of all the characteristics;
step 403, calculate In-Einit,InExpressing the information entropy of the nth feature, taking In-EinitDividing the current sample data set into two subsets by taking the characteristic with the maximum value as a segmentation point; einitRepresenting an initial information entropy;
step 404, recursively calling steps 401 to 403 in the subset part until the entropy of the subset information is 0 or In-EinitWhen the number of the subsets is smaller than the threshold value, stopping continuously dividing the subsets, and finishing the model training;
in the subset at the lowest layer, the type with the largest number of sample types represents the type of the subset;
the artificial intelligence algorithm model returns the type of the subset corresponding to the input features according to the input features.
2. The artificial intelligence analysis method based on transient recording waveforms of claim 1,
after the transient wave recording collector is triggered, the waveform data of 4 periods before the triggering moment and the waveform data of 8 periods after the triggering moment are recorded.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910522198.2A CN110245617B (en) | 2019-06-17 | 2019-06-17 | Artificial intelligence analysis method based on transient recording waveform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910522198.2A CN110245617B (en) | 2019-06-17 | 2019-06-17 | Artificial intelligence analysis method based on transient recording waveform |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110245617A CN110245617A (en) | 2019-09-17 |
| CN110245617B true CN110245617B (en) | 2022-06-24 |
Family
ID=67887546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910522198.2A Active CN110245617B (en) | 2019-06-17 | 2019-06-17 | Artificial intelligence analysis method based on transient recording waveform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110245617B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113033837A (en) * | 2021-03-05 | 2021-06-25 | 国网电力科学研究院武汉南瑞有限责任公司 | Artificial intelligence fault identification system and method based on transient waveform of power transmission line |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105067963A (en) * | 2015-09-24 | 2015-11-18 | 广东电网有限责任公司佛山供电局 | Distribution network fault early warning method and system based on transient waveforms |
| CN107167702A (en) * | 2017-05-04 | 2017-09-15 | 国网福建省电力有限公司 | A kind of distribution feeder fault type recognition method and device |
| CN107329040A (en) * | 2017-06-16 | 2017-11-07 | 国电南瑞科技股份有限公司 | A kind of power distribution automation main station system single-phase earth fault localization method based on transient state recorder data |
| CN107609569A (en) * | 2017-07-31 | 2018-01-19 | 北京映翰通网络技术股份有限公司 | A kind of distribution net work earthing fault localization method based on multidimensional characteristic vectors |
| CN109002762A (en) * | 2018-06-14 | 2018-12-14 | 南方电网科学研究院有限责任公司 | A kind of divide-shut brake coil fault current waveform recognition methods and system |
-
2019
- 2019-06-17 CN CN201910522198.2A patent/CN110245617B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105067963A (en) * | 2015-09-24 | 2015-11-18 | 广东电网有限责任公司佛山供电局 | Distribution network fault early warning method and system based on transient waveforms |
| CN107167702A (en) * | 2017-05-04 | 2017-09-15 | 国网福建省电力有限公司 | A kind of distribution feeder fault type recognition method and device |
| CN107329040A (en) * | 2017-06-16 | 2017-11-07 | 国电南瑞科技股份有限公司 | A kind of power distribution automation main station system single-phase earth fault localization method based on transient state recorder data |
| CN107609569A (en) * | 2017-07-31 | 2018-01-19 | 北京映翰通网络技术股份有限公司 | A kind of distribution net work earthing fault localization method based on multidimensional characteristic vectors |
| CN109002762A (en) * | 2018-06-14 | 2018-12-14 | 南方电网科学研究院有限责任公司 | A kind of divide-shut brake coil fault current waveform recognition methods and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110245617A (en) | 2019-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| CN110398663B (en) | Flexible direct current power grid fault identification method based on convolutional neural network | |
| CN110726898B (en) | Power distribution network fault type identification method | |
| CN111444615B (en) | Photovoltaic array fault diagnosis method based on K nearest neighbor and IV curve | |
| CN110796120A (en) | Time domain feature-based circuit breaker mechanical fault XGboost diagnosis method | |
| Song et al. | A negative selection algorithm-based identification framework for distribution network faults with high resistance | |
| CN115510913B (en) | Fault diagnosis method of H-bridge cascade inverter based on data driving | |
| CN117272143A (en) | Power transmission line fault identification method and device based on gram angle field and residual error network | |
| CN108898182A (en) | A kind of MMC method for diagnosing faults based on core pivot element analysis and support vector machines | |
| CN110703006B (en) | Three-phase power quality disturbance detection method based on convolutional neural network | |
| CN110245617B (en) | Artificial intelligence analysis method based on transient recording waveform | |
| CN111999591A (en) | Method for identifying abnormal state of primary equipment of power distribution network | |
| CN115128345A (en) | Power grid safety early warning method and system based on harmonic monitoring | |
| CN114184870A (en) | Non-invasive load identification method and equipment | |
| CN108108659B (en) | Island detection key feature extraction method based on empirical mode decomposition | |
| CN113702767A (en) | Island direct-current microgrid fault diagnosis method based on wavelet sliding window energy | |
| CN116720095A (en) | Electrical characteristic signal clustering method for optimizing fuzzy C-means based on genetic algorithm | |
| CN114764599B (en) | Power distribution network single-phase earth fault sensitivity analysis method and system | |
| Khond et al. | Data mining methods for bad data detection and event data acquisition in microgrids | |
| CN115879048A (en) | Series arc fault identification method and system based on WRFMDA model | |
| CN115564324A (en) | Electric power fingerprint identification method, device and equipment based on event detection | |
| CN114778969A (en) | Rectifier fault diagnosis method based on RBF-Elman neural network | |
| CN114943254A (en) | A-CNN-based optical fiber differential protection algorithm for output line of photovoltaic power station | |
| CN114252725A (en) | HHT and ResNet 18-based single-phase earth fault type comprehensive identification method | |
| Behzadi et al. | Identification of combined power quality disturbances in the presence of distributed generations using variational mode decomposition and K-nearest neighbors classifier |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |