CN116756660A - Single-phase wire contact vegetation ignition prediction method, system and medium - Google Patents
Single-phase wire contact vegetation ignition prediction method, system and medium Download PDFInfo
- Publication number
- CN116756660A CN116756660A CN202310720826.4A CN202310720826A CN116756660A CN 116756660 A CN116756660 A CN 116756660A CN 202310720826 A CN202310720826 A CN 202310720826A CN 116756660 A CN116756660 A CN 116756660A
- Authority
- CN
- China
- Prior art keywords
- vegetation
- ignition
- current
- fault
- wire contact
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 93
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 45
- 238000007637 random forest analysis Methods 0.000 claims abstract description 33
- 230000001976 improved effect Effects 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000005065 mining Methods 0.000 claims abstract description 10
- 238000003066 decision tree Methods 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 20
- 238000011161 development Methods 0.000 claims description 18
- 238000000354 decomposition reaction Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 11
- 230000008030 elimination Effects 0.000 claims description 11
- 238000003379 elimination reaction Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000010187 selection method Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000005192 partition Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000011426 transformation method Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 description 17
- 238000010586 diagram Methods 0.000 description 16
- 230000008569 process Effects 0.000 description 13
- 238000011156 evaluation Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 238000012544 monitoring process Methods 0.000 description 6
- 230000004044 response Effects 0.000 description 6
- 230000001133 acceleration Effects 0.000 description 5
- 230000006872 improvement Effects 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000002265 prevention Effects 0.000 description 3
- 230000000630 rising effect Effects 0.000 description 3
- WABPQHHGFIMREM-UHFFFAOYSA-N lead(0) Chemical compound [Pb] WABPQHHGFIMREM-UHFFFAOYSA-N 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 235000008331 Pinus X rigitaeda Nutrition 0.000 description 1
- 235000011613 Pinus brutia Nutrition 0.000 description 1
- 241000018646 Pinus brutia Species 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- 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
- 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
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Computer Security & Cryptography (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a single-phase wire contact vegetation ignition prediction method, a system and a medium, comprising the following steps: acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation; performing data feature mining on fault voltage and current signals, and extracting relevant features of the fault voltage and the current; combining the related characteristics of fault voltage and current and vegetation ignition results to form a sample data set; according to the related characteristics of fault voltage and current, carrying out characteristic importance ranking to obtain an important characteristic set and a non-important characteristic set; constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm, and performing model training on the single-phase wire contact vegetation ignition prediction model based on a sample data set; and predicting the ignition condition of the single-phase wire contact vegetation to be detected by adopting a trained single-phase wire contact vegetation ignition prediction model and an important characteristic set. The application has high accuracy in predicting the ignition of the single-phase wire contact vegetation.
Description
Technical Field
The application relates to the technical field of pilot prediction of vegetation in contact with wires, in particular to a pilot prediction method, a pilot prediction system and a pilot prediction medium for vegetation in contact with single-phase wires.
Background
Due to the development of human socioeconomic, the power network passes through a large-area forest coverage area, and in the areas, tree line discharge phenomenon can be generated due to excessively small or direct contact between high and low voltage power transmission and distribution lines and vegetation insulation distance due to windage yaw, broken line falling and the like. In the tree line discharge, the generated heat continuously increases the temperature of branches, when the temperature reaches the ignition point of the branches, the branches are ignited, the power network of China is developed, a plurality of lines pass through the areas with large forest coverage areas, under the extreme weather condition in summer, if the tree line contacts, vegetation can be ignited, forest fires are caused, and the operation safety of the power network is threatened, so that importance is attached to the ignition of the vegetation by the power line to cause the forest fires. If the vegetation ignition risk can be predicted according to the fault signal characteristics, the possibility of forest fire occurrence can be reduced by combining the power line protection strategy. No research on ignition prediction of single-phase single-line contact vegetation has been made in China.
As few domestic researches exist at present, a systematic prediction method does not appear, and the existing similar researches mainly concentrate on the occurrence of wire contact vegetation accidents, and then a fault circuit is larger, so that fault characteristic analysis is carried out for the research of a fault detection algorithm. However, at the initial stage of the wire contact vegetation, the amplitude of a fault signal is weak, and the existing fault characteristics are not suitable for predicting vegetation ignition.
Especially, the prevention of the ignition of the vegetation in contact with the wires mainly depends on patrol observation of power grid patrol personnel in a forest area or whether the ignition phenomenon of the vegetation in contact with the wires is observed through remote monitoring. The existing method has long response time, the accident is found after the vegetation ignition accident occurs due to the contact of the lead wires, however, the vegetation ignition happens at the moment, and the forest fire hazard is high. If in dense vegetation area, manual inspection or remote monitoring, wire contact vegetation accident is difficult to discover.
Disclosure of Invention
The technical problem to be solved by the application is that the existing prevention of the ignition of the vegetation in contact with the wire mainly relies on patrol observation of power grid patrol personnel in a forest area or remote monitoring to observe whether the ignition phenomenon of the vegetation in contact with the wire exists. The existing method has long response time, the accident is found after the vegetation ignition accident caused by the contact of the lead wire, however, the vegetation ignition happens at the moment, and the forest fire hazard is very high. If in dense vegetation area, manual inspection or remote monitoring, wire contact vegetation accident is difficult to discover. In a word, the existing prediction method for the ignition of the vegetation in contact with the single-phase wire is low in accuracy and cannot predict whether the vegetation ignites in advance.
The application aims to provide a single-phase wire contact vegetation ignition prediction method, a single-phase wire contact vegetation ignition prediction system and a single-phase wire contact vegetation ignition prediction medium, which can predict whether vegetation is ignited in advance according to early fault signal characteristics. In the method, vegetation ignition is predicted in the stage of no ignition risk of vegetation when the wires are contacted, so that the aim of early warning is fulfilled. The method has short response time, does not need to observe whether ignition phenomenon exists manually, only needs fault voltage and current signals as data input, does not need to independently preprocess the fault voltage and the current signals, and automatically realizes the extraction of time domain and frequency domain characteristics; meanwhile, by combining an improved random forest algorithm to establish a single-phase wire contact vegetation ignition prediction model, ignition prediction can be performed with high accuracy.
The application is realized by the following technical scheme:
in a first aspect, the present application provides a method for predicting ignition of vegetation in single-phase wire contact, the method comprising:
acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
performing data feature mining on fault voltage and current signals, and extracting relevant features of the fault voltage and the current; the related characteristics comprise fault current time domain instantaneous value characteristics, fault voltage and current time domain development characteristics and fault voltage and current frequency domain characteristics;
combining the related characteristics of fault voltage and current and vegetation ignition results to form a sample data set;
according to the related characteristics of fault voltage and current, carrying out characteristic importance ranking to obtain an important characteristic set and a non-important characteristic set; constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm, and performing model training on the single-phase wire contact vegetation ignition prediction model based on a sample data set;
and predicting the ignition condition of the single-phase wire contact vegetation to be detected by adopting a trained single-phase wire contact vegetation ignition prediction model and an important characteristic set.
Further, the fault voltage and current signals are voltage and current signals when the vegetation is in contact with the high voltage single phase wire in the first stage of vegetation without risk of ignition of the vegetation. This is because the high voltage single phase wire contact vegetation is divided into four phases, and vegetation has been ignited for the second phase, so such voltage and current signals have not been of great significance to vegetation ignition prediction, and so the present application primarily considers the signals in the first phase.
Further, the extraction of the fault current time domain instantaneous value characteristics is as follows: in the time domain, decomposing the fault current signal by adopting an empirical mode decomposition method to obtain the time domain instantaneous value characteristic of the fault current, wherein the method comprises the following steps:
step A1, calculating three equal division points of a fault current signal in a first stage, and taking a plurality of sampling point data at each equal division point;
step B1, performing empirical mode decomposition on a plurality of sampling point data of an equal division point to obtain 8 IMF mode components in total of IMF1-IMF 8;
step C1, after obtaining decomposed IMF modal components, respectively taking {20%,40%,60%,80% } percentile of each IMF modal component as a time domain feature;
and D1, selecting the next equal division point, repeating the step B1 and the step C1, and finishing the feature extraction of the three equal division points to obtain empirical mode decomposition features serving as fault current time domain instantaneous value features.
Further, the extraction of the fault voltage and current time domain development characteristics is as follows:
a2, performing polynomial fitting on fault voltage and current signals to obtain a fitting curve; the highest order of the polynomials is determined by a coefficient R 2 The value satisfies the condition that is greater than the preset value (0.8);
step B2, respectively solving a first derivative and a second derivative of the fitted curve;
and C2, respectively calculating the minimum value min, the maximum value max, the average value mean and the standard deviation std of the first derivative and the second derivative as the time domain development characteristics of the fault voltage and the current.
Further, the extraction of the fault voltage and current frequency domain features is as follows: on the frequency domain, decomposing fault voltage and current signals by adopting a wavelet packet transformation method to obtain fault voltage and current frequency domain characteristics, wherein the method comprises the following steps:
in the frequency domain, after the fault voltage and current signals are subjected to FFT (fast Fourier transform) transformation, a transformation domain result is obtained;
5 layers of decomposition is carried out on the transformation domain by adopting a wavelet packet transformation method, and 32 decomposed nodes are obtained;
from 32 nodes, wavelet packet coefficients of frequency bands where fundamental frequency 50Hz and 3 rd harmonic frequency, 5 th harmonic frequency, 7 th harmonic frequency, 9 th harmonic frequency and 11 th harmonic frequency are located;
extracting wavelet packet coefficients respectively including wavelet packet coefficients and Sum i Mean i Standard deviation Std i Entropy value Entropy i Wavelet packet energy duty cycle E i And characterize them as fault voltage and current frequency domains.
Further, according to the related characteristics of the fault voltage and the fault current, the characteristic importance ranking is performed to obtain an important characteristic set and a non-important characteristic set, including:
according to the related characteristics of the fault voltage and the fault current, calculating the importance of the related characteristics of the fault voltage and the fault current by using a mutual information calculation method, a recursive characteristic elimination method and an embedded characteristic selection method respectively to obtain importance sequences between the related characteristics of the fault voltage and the fault current and vegetation ignition results, namely obtaining a preliminary important characteristic set of each method; and performing intersection operation on the preliminary important feature set obtained by the three methods to obtain an important feature set and a non-important feature set.
Further, the construction steps of the improved random forest algorithm are as follows:
step 1, randomly extracting n samples in a training data sample set D by using an autonomous sampling method Bootstrap Sampling to form a sample subset D k Executing k times in total; wherein, the training data sample set D has n samples and M related feature sets;
step 2, according to the feature importance ranking result of the related features, classifying the related feature set M into: important feature set M 1 Non-significant feature set M 2 ；
Step 3, randomly extracting M in the important feature set M1 1 Features, m 1 ∈[1,M 1 ]The method comprises the steps of carrying out a first treatment on the surface of the At M 2 M is extracted randomly 2 Features, m 2 ＝m-m 1 The method comprises the steps of carrying out a first treatment on the surface of the Therein, whereinRounding;
step 4, constructing a CART decision tree: selecting the minimum Gini coefficient characteristic as the node partition attribute of the decision tree, and repeating the step 4 until no attribute partition or no leaf node sample set can be distinguished;
and 5, repeating the steps 3 to 5, and generating k decision trees to obtain the constructed improved random forest algorithm.
In a second aspect, the present application further provides a single-phase wire contact vegetation ignition prediction system, which uses the single-phase wire contact vegetation ignition prediction method; the system comprises:
the acquisition unit is used for acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
the related characteristic extraction unit is used for carrying out data characteristic mining on fault voltage and current signals and extracting related characteristics of the fault voltage and the current; the related characteristics comprise fault current time domain instantaneous value characteristics, fault voltage and current time domain development characteristics and fault voltage and current frequency domain characteristics;
the sample data set forming unit is used for combining the relevant characteristics of fault voltage and current and vegetation ignition results to form a sample data set;
the feature importance ranking unit is used for ranking the feature importance according to the related features of the fault voltage and the fault current to obtain an important feature set and a non-important feature set;
the prediction model construction and training unit is used for constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm and carrying out model training on the single-phase wire contact vegetation ignition prediction model based on a sample data set;
the ignition prediction unit predicts the ignition condition of the single-phase wire contact vegetation to be detected by adopting a trained single-phase wire contact vegetation ignition prediction model and an important feature set.
Further, the fault voltage and current signals are voltage and current signals when the vegetation is in contact with the high voltage single phase wire in the first stage of vegetation without risk of ignition of the vegetation.
In a third aspect, the present application further provides a computer readable storage medium storing a computer program which when executed by a processor implements a single phase wire contact vegetation ignition prediction method as described above.
Compared with the prior art, the application has the following advantages and beneficial effects:
according to the method, the system and the medium for predicting the ignition of the vegetation in single-phase wire contact, the vegetation ignition is predicted in the stage of no ignition risk of the vegetation in wire contact, and the purpose of early warning is achieved. The method has short response time, does not need to observe whether ignition phenomenon exists manually, only needs fault voltage and current signals as data input, does not need to independently preprocess the fault voltage and the current signals, and automatically realizes the extraction of time domain and frequency domain characteristics; meanwhile, by combining an improved random forest algorithm to establish a single-phase wire contact vegetation ignition prediction model, ignition prediction can be performed with high accuracy.
Drawings
The accompanying drawings, which are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:
FIG. 1 is a flow chart of a single phase wire contact vegetation ignition prediction method of the present application;
FIG. 2 is a schematic diagram of the fault voltage and current signals according to the present application;
FIG. 3 is a second schematic diagram of the fault voltage and current signals of the present application;
FIG. 4 is a flow chart of EMD feature extraction by empirical mode decomposition according to the present application;
FIG. 5 is a schematic diagram of a partial polynomial fit of the present application, (a) fault voltage, (b) fault current;
FIG. 6 is a flow chart of wavelet packet feature extraction according to the present application;
FIG. 7 is a block diagram of a decision tree of the present application;
FIG. 8 is a schematic diagram of a random forest algorithm of the present application;
FIG. 9 is a schematic diagram of the improvement in the improved random forest algorithm of the present application;
FIG. 10 is a schematic diagram of a vegetation ignition prediction model implementation process according to the present application;
FIG. 11 is a graph showing the number of tests and the change of the evaluation index, (a) accuracy, (b) precision, (c) recall, and (d) F1 score;
FIG. 12 is a schematic diagram of the recursive feature elimination principle of the present application;
FIG. 13 is a schematic diagram of the embedded feature selection principle of the present application;
FIG. 14 is a graph showing the results of the mutual information calculation method of the present application;
FIG. 15 is a graph of the result of the recursive feature elimination method of the present application;
FIG. 16 is a graph showing the results of the embedded feature selection method of the present application;
FIG. 17 is a block diagram of a single phase wire contact vegetation ignition prediction system according to the present application.
Detailed Description
For the purpose of making apparent the objects, technical solutions and advantages of the present application, the present application will be further described in detail with reference to the following examples and the accompanying drawings, wherein the exemplary embodiments of the present application and the descriptions thereof are for illustrating the present application only and are not to be construed as limiting the present application.
The existing prevention of the ignition of the vegetation in contact with the wire mainly relies on patrol observation of power grid patrol personnel in a forest area or whether the ignition phenomenon of the vegetation in contact with the wire is observed through remote monitoring. The existing method has long response time, the accident is found after the vegetation ignition accident caused by the contact of the lead wire, however, the vegetation ignition happens at the moment, and the forest fire hazard is very high. If in dense vegetation area, manual inspection or remote monitoring, wire contact vegetation accident is difficult to discover. In a word, the existing prediction method for the ignition of the vegetation in contact with the single-phase wire is low in accuracy and cannot predict whether the vegetation ignites in advance.
Therefore, the application designs a single-phase wire contact vegetation ignition prediction method, a single-phase wire contact vegetation ignition prediction system and a single-phase wire contact vegetation ignition prediction medium, which can predict whether vegetation is ignited in advance according to early fault signal characteristics. In the method, vegetation ignition is predicted in the stage of no ignition risk of vegetation when the wires are contacted, so that the aim of early warning is fulfilled. The method has short response time, does not need to observe whether ignition phenomenon exists manually, only needs fault voltage and current signals as data input, does not need to independently preprocess the fault voltage and the current signals, and automatically realizes the extraction of time domain and frequency domain characteristics; meanwhile, by combining an improved random forest algorithm to establish a single-phase wire contact vegetation ignition prediction model, ignition prediction can be performed with high accuracy.
Example 1
As shown in fig. 1, the method for predicting ignition of vegetation by single-phase wire contact according to the present application comprises:
acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
performing data feature mining on fault voltage and current signals, and extracting relevant features of the fault voltage and the current; the related characteristics comprise fault current time domain instantaneous value characteristics, fault voltage and current time domain development characteristics and fault voltage and current frequency domain characteristics;
combining the related characteristics of fault voltage and current and vegetation ignition results to form a sample data set;
according to the related characteristics of fault voltage and current, carrying out characteristic importance ranking to obtain an important characteristic set and a non-important characteristic set; constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm, and performing model training on the single-phase wire contact vegetation ignition prediction model based on a sample data set;
and predicting the ignition condition of the single-phase wire contact vegetation to be detected by adopting a trained single-phase wire contact vegetation ignition prediction model and an important characteristic set.
The application discloses a single-phase wire contact vegetation ignition prediction method which comprises the steps of fault voltage and current data acquisition and analysis, data characteristic mining, matlab-based prediction algorithm development, prediction model establishment, method verification and the like.
1. Fault voltage and current signals
The multiple sets of test data were analyzed and the fault current signal was found to exhibit significant 4 phases, as shown in fig. 2. In combination with experimental phenomenon analysis, in the first stage of the high-voltage wire contacting vegetation, the vegetation has no ignition risk, so that the first stage fault voltage and current are used as research stages to develop signal characteristic excavation.
As shown in fig. 3, a section of fault voltage and current waveforms in the vegetation ignition process by wire contact have distortion near voltage and current peaks by observing the change of fault signals, and have larger difference from a normal sine waveform. The fault voltage and the fault current have rich characteristic information, and signal diagnosis is mined from the time domain and the frequency domain respectively.
Therefore, the fault voltage and current signal of the present application is the voltage and current signal when the vegetation is in the first stage of the high voltage single phase wire contact vegetation without ignition risk. This is because the high voltage single phase wire contact vegetation is divided into four phases, and vegetation has been ignited for the second phase, so such voltage and current signals have not been of great significance to vegetation ignition prediction, and so the present application primarily considers the signals in the first phase.
2. Data feature mining
Performing data feature mining on fault voltage and current signals, and extracting relevant features of the fault voltage and the current; the related characteristics comprise fault current time domain instantaneous value characteristics, fault voltage and current time domain development characteristics and fault voltage and current frequency domain characteristics;
1, time domain instantaneous value characteristics of fault current
The fault current signal is decomposed in the time domain using an empirical mode decomposition (Empirical Mode Decomposition, EMD). EMD is used to analyze non-stationary, nonlinear signals. The EMD decomposes the complex signal into a plurality of finite modality functions (Intrinsic Mode Functions, IMFs) and residual terms (Residue functions), each IMF occupying a frequency band.
Specifically, the empirical mode decomposition and feature extraction process is shown in fig. 4:
step A1, calculating three equal division points of a fault current signal in a first stage, and taking 1000 sampling point data at each equal division point;
step B1, performing empirical mode decomposition on a plurality of sampling point data of an equal division point to obtain 8 IMF mode components in total of IMF1-IMF 8;
step C1, after obtaining decomposed IMF modal components, respectively taking {20%,40%,60%,80% } percentile of each IMF modal component as a time domain feature;
and D1, selecting the next equal division point, repeating the step B1 and the step C1, and finishing the feature extraction of the three equal division points to obtain empirical mode decomposition features serving as fault current time domain instantaneous value features.
According to the above steps A1 to D1, 96 empirical mode decomposition features are obtained in total, i.e. 96 fault current time domain instantaneous value features are obtained in total.
2 time domain development characteristics of fault voltage and current
Fitting a local polynomial on fault voltage and current signals in the first stage by using a Matlab curve fitting tool box, and determining a coefficient R by using a fitting error 2 (R-Square) measurement, coefficient R is determined 2 The value range is 0,1]The larger the value, the better the fitting effect, and the polynomial fitting degree is determined by a coefficient R in order to ensure the curve fitting effect due to the difference between the fault voltage and the current waveforms of each group 2 The value satisfies a conditional determination of greater than 0.8.
And solving the first derivative and the second derivative of the polynomial according to the fitted curve to obtain the voltage falling rate, the falling acceleration, the current rising rate and the rising acceleration. As shown in fig. 5, the fitted feature extraction is performed by taking pine fault voltage and current as an example.
Taking the first derivative and the second derivative of the fitting curve, as follows:
in formulas (5) - (6), I is current; u is voltage; v (V) I Is the rate of rise of the current; a, a I Is the current rising acceleration; v (V) U Is the rate of decrease of the current; a, a U Is the voltage drop acceleration.
And respectively taking a minimum value min, a maximum value max, an average value mean and a standard deviation std for the voltage and current change rates and the acceleration.
Specifically, the fault voltage and current time domain development characteristic extraction steps are as follows:
a2, performing polynomial fitting on fault voltage and current signals to obtain a fitting curve; the highest order of the polynomials is determined by a coefficient R 2 A value satisfying a condition determination greater than 0.8;
step B2, respectively solving a first derivative and a second derivative of the fitted curve;
and C2, respectively calculating the minimum value min, the maximum value max, the average value mean and the standard deviation std of the first derivative and the second derivative as the time domain development characteristics of the fault voltage and the current.
According to the steps A2 to C2, 12 fault voltage and current time domain development characteristics are obtained.
3, fault voltage and current frequency domain characteristics
On the frequency domain, decomposing fault voltage and current signals by adopting a wavelet packet transformation method to obtain fault voltage and current frequency domain characteristics; comprising the following steps:
after the fault voltage and current signals are subjected to FFT (fast Fourier transform) transformation, odd harmonics such as 3, 5, 7, 9, 11 and the like in the frequency content have obvious amplitude values, the fault voltage and current signals have abundant frequency domain characteristics, and the fault signals are decomposed by wavelet packet transformation to obtain high-frequency coefficients and low-frequency coefficients of the signals for extracting the frequency domain information.
After decomposing the signal, extracting wavelet packet characteristics of 6 nodes with fundamental frequency of 50Hz and 3, 5, 7, 9 and 11 harmonic frequencies, including wavelet packet coefficients and Sum i Mean i Standard deviation Std i Entropy value Entropy i Wavelet packet energy duty cycle E i . The extraction process is shown in fig. 6:
(1) And carrying out wavelet packet 5-layer decomposition on the fault voltage and current signals in the first stage to obtain 32 nodes.
(2) And acquiring wavelet packet coefficients of frequency bands where 50Hz,3, 5, 7, 9 and 11 subharmonic frequencies are located.
(3) Extracting wavelet packet coefficients respectively including wavelet packet coefficients and Sum i Mean i Standard deviation Std i Entropy value Entropy i Wavelet packet energy duty cycle E i And characterizing them as fault voltage and current frequency domains.
The above frequency domain features are extracted from the fault voltage and current signals in the first stage respectively, 60 frequency domain features can be obtained, and 60+96+12=168 features are combined with the time domain features.
3. Matlab prediction algorithm based development
The random forest algorithm is an integrated learning type in machine learning, and is widely applied to classification problems due to the simple random forest parameter setting and strong generalization capability. The integrated learning is to train a plurality of weak classification learners to obtain a strong classification learner, and the weak classification learners in the random forest are decision trees. The decision tree structure is shown in fig. 7, and the process from the initial root node to the final leaf node is a branch node generated by dividing different characteristic attributes. The classification process is to divide different characteristic attributes of the sample data, and finally classify the sample into a leaf node, so as to realize classification of the sample.
The random forest algorithm generates a plurality of CART decision trees by combining with the Bagging algorithm, and the classification results of the decision trees are combined, so that the judgment of sample categories is realized, the classification accuracy and the generalization effect of the algorithm on sample data are improved by the integrated learning method, and the algorithm principle is shown in figure 8.
The application improves the random forest algorithm to obtain higher prediction performance, and according to the thought of feature hierarchical sampling, the application firstly adopts a mutual information calculation method to evaluate the importance of the features. The mutual information method is a useful information measure in information theory, which can be seen as the amount of information contained in one random variable about another, and can be taken as the importance of the sample feature.
After the feature importance ranking is obtained, the features are divided into important feature sets M 1 Non-important feature set M 2 Extracting feature subsets by adopting a random sampling method in the process of generating a decision tree, and ensuring that the feature subsets are over-sampled in the process of generating the decision treeIn the process, each decision tree has important characteristics, and the algorithm improvement content is shown in fig. 9.
The improvement specific steps of the random forest feature selection method are as follows:
(1) Firstly, important feature subset M is obtained according to feature importance ranking 1 The rest of the features are taken as non-important feature sets M 2 。
(2) Adopting a non-repeated random selection method in the important feature set M 1 Selecting m 1 ∈[1,M 1 ]Features in the non-important feature set M 2 In selecting m 2 And features.
(3) Selecting a feature subset m as a feature set of a generating decision tree, wherein the feature number m=m 1 +m 2 。
Through the steps, each decision tree has important characteristics in the decision tree generation process, and the number of the important characteristics is random, so that the accuracy of the decision tree is ensured, meanwhile, the diversity of the decision tree in a random forest is not reduced, and the overall classification performance is improved.
The Matlab is utilized for software development, and the construction steps of the improved random forest algorithm are as follows:
step 1, randomly extracting n samples in a training data sample set D by using an autonomous sampling method Bootstrap Sampling to form a sample subset D k Executing k times in total; wherein, the training data sample set D has n samples and M related feature sets;
step 2, according to the feature importance ranking result of the related features, classifying the related feature set M into: important feature set M 1 Non-significant feature set M 2 ；
Step 3, in the important feature set M 1 M is extracted randomly 1 Features, m 1 ∈[1,M 1 ]The method comprises the steps of carrying out a first treatment on the surface of the At M 2 M is extracted randomly 2 Features, m 2 ＝m-m 1 The method comprises the steps of carrying out a first treatment on the surface of the Therein, whereinRounding;
step 4, constructing a CART decision tree: selecting the minimum Gini coefficient characteristic as the node partition attribute of the decision tree, and repeating the step 4 until no attribute partition or no leaf node sample set can be distinguished;
and 5, repeating the steps 3 to 5, and generating k decision trees to obtain the constructed improved random forest algorithm.
As a further implementation, the feature importance ranking is performed according to the relevant features of the fault voltage and the fault current, to obtain an important feature set and a non-important feature set, including:
to obtain the important feature set M 1 、M 2 The importance of the sample characteristics is required to be evaluated, and the importance of the relevant characteristics of the fault voltage and the current is calculated by a mutual information calculation method, a recursive characteristic elimination method and an embedded characteristic selection method respectively according to the relevant characteristics of the fault voltage and the current, so that the importance ranking between the relevant characteristics of the fault voltage and the current and the vegetation ignition result is obtained, namely, a preliminary important characteristic set of each method is obtained; and performing intersection operation on the preliminary important feature set obtained by the three methods to obtain an important feature set and a non-important feature set.
(1) Mutual information calculation method
Mutual information computation is a useful information metric in information theory, which can be seen as the amount of information contained in one random variable about another random variable. Two random variables are respectively x and y, the corresponding edge probability distribution is P (x) and P (y), the joint probability distribution is P (x, y), and the following mutual information calculation formula between x and y is as follows:
the larger the value of I (x, y), the greater the correlation between the features x and y, and if x and y are independent of each other, I (x, y) is 0. By calculating I (x, y), the correlation between the voltage-current characteristics and the ignition result is obtained.
(2) Recursive feature elimination method
The recursive feature elimination method is to iterate for a plurality of times based on a specified model, calculate the importance of the current feature set according to an algorithm in each iteration process, and sort the importance. And eliminating the features with small importance in each iteration, so that the feature set is smaller and smaller until the feature number reaches the target set number, and stopping the iteration. The decision tree is selected as a reference model of iteration pairs, important features are extracted from the final model, and the principle of recursive feature elimination is shown in fig. 12. The recursive feature elimination steps are as follows:
1) Selecting a decision tree as a reference model;
2) Taking sample data as input, including multidimensional data features and sample categories;
3) In each iteration, the algorithm model evaluates the variable weight and deletes the lowest importance feature;
4) Step 3 is repeated until a specified number of features are reached. The retained features are finally selected as input features for the final model.
(3) Embedded feature selection method
The embedded feature selection is to evaluate the sample features during the algorithm training process, and the feature evaluation is performed synchronously with the model, as shown in fig. 13.
Because the method performs feature screening by the model itself, the obtained important features have great influence on the model. The embedded feature selection steps are as follows:
1) The predictive error before the jth feature of the sample is not interfered can be obtained by utilizing the test data in the random forest algorithm 1 (j)。
2) Randomly adding interference to the jth feature in the test data set to obtain a prediction error 2 (j)。
3) And calculating a difference delta error (j) of the two errors, and evaluating the difference delta error (j) by using the importance of the delta error (j) characteristic j.
If the features are important, the larger Δerror (j); if the test error Δerror (j) varies little, it is stated that feature j is significantly low. Δerror (j) is:
Δerror(j)＝|error 1 (j)-error 2 (j)| (6)
the importance of the voltage and current characteristics is obtained by calculating the feature set by using the above three methods, and the calculation results of the three methods are shown in fig. 14, 15 and 16.
The feature importance is ordered by three methods of fusion mutual information calculation method, recursion feature elimination method and embedded feature selection method, and intersection of the three methods is taken as an important feature set M 1 The remaining features are taken as non-important feature sets M 2 . Proved by experiments, when M 1 At a number of 6, the test accuracy of the model is highest, so M will be 1 Set to 6.
4. Prediction model establishment
And combining the proposed improved random forest algorithm, establishing a single-phase wire contact vegetation ignition prediction model, and realizing the flow as shown in figure 10. The vegetation ignition prediction process may be divided into the following steps:
(1) And taking fault voltage and current signals of multiple groups of ignition tests and non-ignition tests as data input of a model.
(2) Through wavelet packet transformation, EMD decomposition obtains time domain and frequency domain characteristics of fault voltage and current.
(3) And combining the fault signal characteristics and the vegetation ignition result to form a sample data set.
(4) And evaluating the importance of the feature evaluation by using three methods, namely a mutual information calculation method, a recursive feature elimination method and an embedded feature selection method.
(5) Training an improved random forest algorithm by using a training set, predicting whether vegetation is ignited or not according to the characteristic input of a test sample, and establishing a vegetation ignition prediction model.
5. Method verification
1. Vegetation ignition test verification
The test resulted in 321 sets of effective test data, 121 of which were non-ignited and 200 of which were ignited. A total of 168 voltage and current characteristics were included, for a total of 168×311= 52248 data. Comprehensively considering 4 evaluation indexes of accuracy, precision, recall rate and F1 fraction value in prediction classification, and verifying the single-phase wire contact vegetation ignition prediction method. To eliminate the deviation of the algorithm predicted results, 20% of the test data were randomly extracted as test sets each time, with the pilot and non-pilot groups each accounting for half. Each algorithm was tested 100 times to obtain the number of tests and each evaluation index change, as shown in fig. 11.
From fig. 11, it can be seen that each evaluation index of the single-phase wire contact vegetation ignition prediction model has better performance, and the minimum value, the maximum value, the mean value and the standard deviation are counted in the figure. The average value of the prediction accuracy of the single-phase lead contact vegetation ignition prediction model on vegetation ignition results reaches 92.67%, the maximum accuracy reaches 96.59%, the discrimination between vegetation ignition and non-ignition results is high, the average value of the accuracy is 96.20%, and the maximum is 100%; the recall rate of other indexes and the average value of the F1 score value reach 89.70 percent and 92.78 percent respectively.
2. Algorithm comparison
In order to compare the improved effect of the algorithm, the test result is compared with a random forest algorithm, a Support Vector Machine (SVM), an Ada Boost, a nearest neighbor classification algorithm (KNN) and a gradient lifting algorithm, and the average value of the 4 evaluation indexes is shown in table 1.
Table 1 algorithm comparison results
The comparison result shows that compared with the random forest algorithm, the improved random forest algorithm has obvious performance improvement, and the average value of the accuracy, the precision and the F1 fraction value is higher than that of the random forest algorithm.
The test verification result proves that the single-phase wire contact vegetation ignition prediction method has higher accuracy and effectiveness, and can perform ignition prediction according to the fault voltage and current signal characteristics before vegetation ignition.
In the application, the fault voltage and current signal characteristics are extracted at the risk stage of no ignition of the vegetation contacted by the lead; after fault voltage and current signals are input, the fault signals do not need to be preprocessed independently, and the time domain and frequency domain characteristics are extracted automatically; and the fault voltage and current signals are used as data input, the influence of complex factors such as environmental temperature and humidity, vegetation height, types, water content and the like is attributed to the influence on the characteristics of the fault signals, and the prediction of the ignition of the vegetation in contact with the wires is realized only by means of the early fault voltage and current of the vegetation in contact with the wires. The improved random forest algorithm is introduced to establish the single-phase wire contact vegetation ignition prediction model, so that the method has higher accuracy.
Example 2
As shown in fig. 17, the present embodiment is different from embodiment 1 in that this embodiment further provides a single-phase wire contact vegetation ignition prediction system using a single-phase wire contact vegetation ignition prediction method of embodiment 1; the system comprises:
the acquisition unit is used for acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
the related characteristic extraction unit is used for carrying out data characteristic mining on fault voltage and current signals and extracting related characteristics of the fault voltage and the current; the related characteristics comprise fault current time domain instantaneous value characteristics, fault voltage and current time domain development characteristics and fault voltage and current frequency domain characteristics;
the sample data set forming unit is used for combining the relevant characteristics of fault voltage and current and vegetation ignition results to form a sample data set;
the feature importance ranking unit is used for ranking the feature importance according to the related features of the fault voltage and the fault current to obtain an important feature set and a non-important feature set;
the prediction model construction and training unit is used for constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm and carrying out model training on the single-phase wire contact vegetation ignition prediction model based on a sample data set;
the ignition prediction unit predicts the ignition condition of the single-phase wire contact vegetation to be detected by adopting a trained single-phase wire contact vegetation ignition prediction model and an important feature set.
Further, the fault voltage and current signals are voltage and current signals when the vegetation is in contact with the high voltage single phase wire in the first stage of vegetation without risk of ignition of the vegetation.
The execution process of each unit is performed according to the steps of the single-phase wire contact vegetation ignition prediction method in embodiment 1, and the details of this embodiment are not repeated.
Meanwhile, the present application also provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements a single-phase wire contact vegetation ignition prediction method of embodiment 1.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the application, and is not meant to limit the scope of the application, but to limit the application to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the application are intended to be included within the scope of the application.
Claims (10)
1. A single-phase wire contact vegetation ignition prediction method, which is characterized by comprising the following steps:
acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
performing data feature mining on the fault voltage and current signals, and extracting relevant features of the fault voltage and the fault current; the related characteristics comprise a fault current time domain instantaneous value characteristic, a fault voltage and current time domain development characteristic and a fault voltage and current frequency domain characteristic;
combining the related characteristics of the fault voltage and the fault current with vegetation ignition results to form a sample data set;
according to the related characteristics of the fault voltage and the fault current, carrying out characteristic importance ranking to obtain an important characteristic set and a non-important characteristic set; constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm, and performing model training on the single-phase wire contact vegetation ignition prediction model based on the sample data set;
and predicting the ignition condition of the single-phase wire contact vegetation to be detected by adopting the trained single-phase wire contact vegetation ignition prediction model and the important feature set.
2. The method of claim 1, wherein the fault voltage and current signal is a voltage and current signal when vegetation is not at risk for ignition during a first phase of high voltage single phase wire contact vegetation.
3. The single-phase wire contact vegetation ignition prediction method according to claim 1, wherein the extraction of the fault current time domain instantaneous value characteristics is: in the time domain, decomposing the fault current signal by adopting an empirical mode decomposition method to obtain the time domain instantaneous value characteristic of the fault current, wherein the method comprises the following steps:
step A1, calculating three equal division points of a fault current signal, and taking a plurality of sampling point data at each equal division point;
step B1, performing empirical mode decomposition on a plurality of sampling point data of an equal division point to obtain 8 IMF mode components in total of IMF1-IMF 8;
step C1, after obtaining decomposed IMF modal components, respectively taking {20%,40%,60%,80% } percentile of each IMF modal component as a time domain feature;
and D1, selecting the next equal division point, repeating the step B1 and the step C1, and finishing the feature extraction of the three equal division points to obtain empirical mode decomposition features serving as fault current time domain instantaneous value features.
4. The method for predicting ignition of single-phase wire contact vegetation according to claim 1, wherein the extracting of the fault voltage and current time domain development features is as follows:
a2, performing polynomial fitting on the fault voltage and current signals to obtain a fitting curve; the highest order of the polynomial is determined by the condition that the decision coefficient meets a preset value;
step B2, respectively solving a first derivative and a second derivative of the fitting curve;
and C2, respectively calculating the minimum value, the maximum value, the average value and the standard deviation of the first derivative and the second derivative as time domain development characteristics of fault voltage and current.
5. The single-phase wire contact vegetation ignition prediction method of claim 1, wherein the extraction of the fault voltage and current frequency domain features is: on the frequency domain, decomposing the fault voltage and current signals by adopting a wavelet packet transformation method to obtain the fault voltage and current frequency domain characteristics, comprising the following steps:
in the frequency domain, the fault voltage and current signals are subjected to FFT (fast Fourier transform) to obtain a transform domain result;
performing 5-layer decomposition on the transformation domain by adopting a wavelet packet transformation method to obtain 32 decomposed nodes;
from the 32 nodes, wavelet packet coefficients of frequency bands where fundamental frequency 50Hz and 3 rd harmonic frequency, 5 th harmonic frequency, 7 th harmonic frequency, 9 th harmonic frequency and 11 th harmonic frequency are located;
and respectively extracting wavelet packet coefficients including wavelet packet coefficient sum, mean value, standard deviation, entropy value and wavelet packet energy duty ratio, and taking the wavelet packet coefficients as fault voltage and current frequency domain characteristics.
6. The method for predicting ignition of single-phase wire contact vegetation according to claim 1, wherein the ranking of feature importance according to the relevant features of the fault voltage and the fault current to obtain an important feature set and a non-important feature set comprises:
according to the related characteristics of the fault voltage and the fault current, calculating the importance of the related characteristics of the fault voltage and the fault current by a mutual information calculation method, a recursive characteristic elimination method and an embedded characteristic selection method respectively to obtain importance sequences between the related characteristics of the fault voltage and the fault current and vegetation ignition results, namely obtaining a preliminary important characteristic set of each method; and performing intersection operation on the preliminary important feature set obtained by the three methods to obtain an important feature set and a non-important feature set.
7. The method for predicting ignition of single-phase wire contact vegetation according to claim 1, wherein the construction steps of the improved random forest algorithm are as follows:
step 1, randomly extracting n samples in a training data sample set D by using an autonomous sampling method to form a sample subset D k Executing k times in total; wherein, the training data sample set D has n samples and M related feature sets;
step 2, according to the feature importance ranking result of the related features, classifying the related feature set M into: important feature set M 1 Non-significant feature set M 2 ；
Step 3, in the important feature set M 1 M is extracted randomly 1 Features, m 1 ∈[1,M 1 ]The method comprises the steps of carrying out a first treatment on the surface of the At M 2 M is extracted randomly 2 Features, m 2 ＝m-m 1 The method comprises the steps of carrying out a first treatment on the surface of the Therein, whereinRounding;
step 4, constructing a CART decision tree: selecting the minimum Gini coefficient characteristic as the node partition attribute of the decision tree, and repeating the step 4 until no attribute partition or no leaf node sample set can be distinguished;
and 5, repeating the steps 3 to 5, and generating k decision trees to obtain the constructed improved random forest algorithm.
8. A single phase wire contact vegetation ignition prediction system, the system comprising:
the acquisition unit is used for acquiring fault voltage and current signals of a plurality of groups of single-phase wire contact vegetation;
the related characteristic extraction unit is used for carrying out data characteristic mining on the fault voltage and current signals and extracting related characteristics of the fault voltage and the current; the related characteristics comprise a fault current time domain instantaneous value characteristic, a fault voltage and current time domain development characteristic and a fault voltage and current frequency domain characteristic;
the sample data set forming unit is used for combining the relevant characteristics of the fault voltage and the fault current and vegetation ignition results to form a sample data set;
the feature importance ranking unit is used for ranking the feature importance according to the related features of the fault voltage and the fault current to obtain an important feature set and a non-important feature set;
the prediction model construction and training unit is used for constructing a single-phase wire contact vegetation ignition prediction model by adopting an improved random forest algorithm and carrying out model training on the single-phase wire contact vegetation ignition prediction model based on the sample data set;
and the ignition prediction unit predicts the ignition condition of the single-phase wire contact vegetation to be detected by adopting the trained single-phase wire contact vegetation ignition prediction model and the important feature set.
9. The single phase wire contact vegetation ignition prediction system of claim 8 wherein the fault voltage and current signal is a voltage and current signal when vegetation is at risk of ignition during the first phase of high voltage single phase wire contact vegetation.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a single phase wire contact vegetation ignition prediction method as claimed in any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310720826.4A CN116756660B (en) | 2023-06-16 | 2023-06-16 | Single-phase wire contact vegetation ignition prediction method, system and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310720826.4A CN116756660B (en) | 2023-06-16 | 2023-06-16 | Single-phase wire contact vegetation ignition prediction method, system and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116756660A true CN116756660A (en) | 2023-09-15 |
| CN116756660B CN116756660B (en) | 2024-02-06 |
Family
ID=87947361
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310720826.4A Active CN116756660B (en) | 2023-06-16 | 2023-06-16 | Single-phase wire contact vegetation ignition prediction method, system and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116756660B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005304114A (en) * | 2004-04-07 | 2005-10-27 | Hitachi Ltd | Tree contact monitor of distribution line |
| KR20210150775A (en) * | 2020-06-04 | 2021-12-13 | 한국전력공사 | Apparatus and method for cutting off distribution line |
| CN114764599A (en) * | 2022-04-26 | 2022-07-19 | 国网四川省电力公司电力科学研究院 | Sensitivity analysis method and system for single-phase earth fault of power distribution network |
| CN115017444A (en) * | 2022-04-18 | 2022-09-06 | 上海交通大学 | Positioning and identifying method for early faults of tree line |
| CN115113092A (en) * | 2022-03-31 | 2022-09-27 | 云南电网有限责任公司电力科学研究院 | Tree line early fault feature extraction method, live-action simulation equipment, system, computer equipment and medium |
| US20220373612A1 (en) * | 2019-11-06 | 2022-11-24 | Newsouth Innovations Pty Limited | Apparatus and process for real-time detection of high-impedance faults in power lines |
| US20230012038A1 (en) * | 2019-11-27 | 2023-01-12 | Sentient Technology Holdings, LLC | Systems and methods for automated detection of switch capacitor operation |
| CN115639442A (en) * | 2022-10-19 | 2023-01-24 | 国网福建省电力有限公司泉州供电公司 | Method and system for identifying tree line contradiction discharge fault time sequence of medium-voltage line |
| CN115758634A (en) * | 2022-10-25 | 2023-03-07 | 昆明能讯科技有限责任公司 | Method for initiating tree combustion process space-time evolution through circuit discharge and storage medium |
| CN115905910A (en) * | 2022-09-26 | 2023-04-04 | 上海交通大学 | Distribution line tree line early fault generation mechanism and identification method |
-
2023
- 2023-06-16 CN CN202310720826.4A patent/CN116756660B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005304114A (en) * | 2004-04-07 | 2005-10-27 | Hitachi Ltd | Tree contact monitor of distribution line |
| US20220373612A1 (en) * | 2019-11-06 | 2022-11-24 | Newsouth Innovations Pty Limited | Apparatus and process for real-time detection of high-impedance faults in power lines |
| US20230012038A1 (en) * | 2019-11-27 | 2023-01-12 | Sentient Technology Holdings, LLC | Systems and methods for automated detection of switch capacitor operation |
| KR20210150775A (en) * | 2020-06-04 | 2021-12-13 | 한국전력공사 | Apparatus and method for cutting off distribution line |
| CN115113092A (en) * | 2022-03-31 | 2022-09-27 | 云南电网有限责任公司电力科学研究院 | Tree line early fault feature extraction method, live-action simulation equipment, system, computer equipment and medium |
| CN115017444A (en) * | 2022-04-18 | 2022-09-06 | 上海交通大学 | Positioning and identifying method for early faults of tree line |
| CN114764599A (en) * | 2022-04-26 | 2022-07-19 | 国网四川省电力公司电力科学研究院 | Sensitivity analysis method and system for single-phase earth fault of power distribution network |
| CN115905910A (en) * | 2022-09-26 | 2023-04-04 | 上海交通大学 | Distribution line tree line early fault generation mechanism and identification method |
| CN115639442A (en) * | 2022-10-19 | 2023-01-24 | 国网福建省电力有限公司泉州供电公司 | Method and system for identifying tree line contradiction discharge fault time sequence of medium-voltage line |
| CN115758634A (en) * | 2022-10-25 | 2023-03-07 | 昆明能讯科技有限责任公司 | Method for initiating tree combustion process space-time evolution through circuit discharge and storage medium |
Non-Patent Citations (4)
| Title |
|---|
| GOMES D P S: "《Vegetation highimpedance faults’high-frequency signatures via sparse coding》", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》, vol. 69, no. 7 * |
| MA JUN: "《Real-time detection of wildfire risk caused by powerline vegetation faults using advanced machine learning techniques》", 《ADVANCED ENGINEERING INFORMATICS》 * |
| NING XIN: "《Detection Method of 10kV Overhead Line SinglePhase Contact Tree Grounding Fault Based on MultiResolution Wavelet Transform and Multi-Parameter Information Fusion》", 《HTTPS://IEEEXPLORE.IEEE.ORG/STAMP/STAMP.JSP?TP=&ARNUMBER=10140978》, pages 1 - 6 * |
| 秦际轲: "《架空导线单相触树接地故障的放电引燃机制研究》", 《电网技术》, no. 3 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116756660B (en) | 2024-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107101813B (en) | A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal | |
| Peng et al. | Random forest based optimal feature selection for partial discharge pattern recognition in HV cables | |
| CN107451557B (en) | Power transmission line short-circuit fault diagnosis method based on empirical wavelet transform and local energy | |
| CN103076547B (en) | Method for identifying GIS (Gas Insulated Switchgear) local discharge fault type mode based on support vector machines | |
| CN105067966B (en) | The low-voltage alternating-current fault arc detection method of feature based modal components energy spectrometer | |
| CN108008244B (en) | A kind of small current grounding fault progressive classifying identification method at many levels | |
| CN105320969A (en) | A heart rate variability feature classification method based on multi-scale Renyi entropy | |
| Ray et al. | Modified wavelet transform based fault analysis in a solar photovoltaic system | |
| CN114325256A (en) | Power equipment partial discharge identification method, system, equipment and storage medium | |
| CN114004091A (en) | CEEMDAN-BNs-based wind variable pitch system fault diagnosis method | |
| CN112486137A (en) | Method and system for constructing fault feature library of active power distribution network and fault diagnosis method | |
| CN115079052A (en) | Transformer fault diagnosis method and system | |
| CN116756660B (en) | Single-phase wire contact vegetation ignition prediction method, system and medium | |
| Nicolae et al. | Analyzing Electromagnetic Interferences in Power Applications by Using Time-Efficient Joint Analysis Based on DWT and WPT Trees | |
| Xu et al. | A vibration signal anomaly detection method based on frequency component clustering and isolated forest algorithm | |
| CN113866614A (en) | Multi-scenario user side low-voltage direct-current switch arc fault diagnosis method and device | |
| CN110703013B (en) | Online identification method and device for low-frequency oscillation mode of power system and electronic equipment | |
| CN116298735A (en) | AC arc fault detection method and related device for low-voltage distribution network | |
| CN115902557A (en) | Switch cabinet fault diagnosis processing method and device and nonvolatile storage medium | |
| CN116031828A (en) | 10kV cable early fault identification method based on wavelet transformation and random forest algorithm | |
| CN112729531B (en) | Fault studying and judging method and system for distribution transformer equipment | |
| CN115327325A (en) | Charging pile series arc fault detection method and device and electronic equipment | |
| Anshuman et al. | A novel hybrid algorithm for event detection, localisation and classification | |
| CN110703080B (en) | GIS spike discharge diagnosis method, discharge degree identification method and device | |
| CN114091593A (en) | Network-level arc fault diagnosis method based on multi-scale feature fusion |
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 |