CN113608082B - Ring main unit discharge state identification method based on audio signals - Google Patents
Ring main unit discharge state identification method based on audio signals Download PDFInfo
- Publication number
- CN113608082B CN113608082B CN202110868613.7A CN202110868613A CN113608082B CN 113608082 B CN113608082 B CN 113608082B CN 202110868613 A CN202110868613 A CN 202110868613A CN 113608082 B CN113608082 B CN 113608082B
- Authority
- CN
- China
- Prior art keywords
- main unit
- ring main
- energy
- sound signal
- unit discharge
- 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
- 230000005236 sound signal Effects 0.000 title claims abstract description 134
- 238000000034 method Methods 0.000 title claims abstract description 48
- 239000013598 vector Substances 0.000 claims abstract description 96
- 230000009467 reduction Effects 0.000 claims abstract description 51
- 238000010977 unit operation Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 8
- 238000007599 discharging Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000012706 support-vector machine Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 239000012212 insulator Substances 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 5
- 238000009413 insulation Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 8
- 230000007547 defect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1209—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing using acoustic measurements
Landscapes
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Testing Relating To Insulation (AREA)
Abstract
The invention relates to the technical field of ring main unit fault identification, and particularly discloses a ring main unit discharge state identification method based on audio signals, which comprises the following steps: collecting ring main unit operation sound signals in different states, wherein the ring main unit operation sound signals in different states comprise ring main unit discharge sound signals in different types; processing the ring main unit operation sound signal to obtain an energy characteristic vector of the ring main unit discharge sound signal; the dimension of the energy feature vector of the ring main unit discharge sound signal is reduced by using the Fisher dimension reduction criterion, and the dimension-reduced energy feature vector is obtained; and establishing a ring main unit discharge sound signal identification model, and inputting the energy feature vector after the dimension reduction into the ring main unit discharge sound signal identification model to obtain a ring main unit discharge sound signal identification result. The method and the device can improve the identification precision of the faults of the ring main unit, are quicker and more reliable, and have great guiding effect on timely finding the faults.
Description
Technical Field
The invention relates to the technical field of ring main unit fault identification, in particular to a ring main unit discharge state identification method based on audio signals.
Background
The ring main unit is used as an important central link for control and protection in the power system, and becomes a strong rear shield for ensuring the stable operation of the power distribution network. During the manufacturing and operation of the electric power grid, partial discharge can be caused by various defects, and the stable operation of the electric power grid is threatened.
Aiming at the fault problem of the ring main unit, real-time control is required, measures are taken to avoid the expansion of the situation, and the state information of the equipment is required to be acquired. There are a number of disadvantages to relying on the traditional way that a worker listens to and looks at through the ear to discern whether the device is in good condition. Meanwhile, the traditional manual method cannot predict the state of the equipment at the next moment, and cannot predict potential faults. With the proposal of smart grids, the structure of electrical devices is more and more complex. When faults occur, various faults have strong randomness and diversity. Thus, applying modern technology in the field of digital signal processing, the required data information can be obtained by means of more advanced technology.
In the failure of the ring main unit, the failure is mainly an insulation defect. The quality of the insulation state is directly related to the safe operation of the electrical equipment and the lines. In recent years, power failure disasters have caused large losses to national economy. To minimize losses, a reliable set of insulation monitoring methods must be developed. Partial discharge is one of the important forms of monitoring the insulation state, and can be monitored by establishing a reasonable monitoring system.
During operation of the high voltage device, partial discharges are often accompanied by some audible sound. Experiments and related researches find that the sounds have a certain identification property, and the insulation state can be judged by distinguishing the characteristics of the sounds. The invention provides a ring main unit discharge state identification method based on an audio signal by selecting sound as an identification object.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a ring main unit discharge state identification method based on audio signals, which is used for carrying out sound on-line monitoring on insulation faults of the ring main unit and guaranteeing safe and reliable operation of the ring main unit.
As a first aspect of the present invention, there is provided an audio signal-based ring main unit discharge state identification method, including:
step S1: collecting ring main unit operation sound signals in different states, wherein the ring main unit operation sound signals in different states comprise ring main unit discharge sound signals in different types;
step S2: processing the ring main unit operation sound signal to obtain an energy characteristic vector of a ring main unit discharge sound signal, wherein the energy characteristic vector of the ring main unit discharge sound signal represents the characteristic of a discharge sound signal waveform;
step S3: reducing the dimension of the energy feature vector of the ring main unit discharge sound signal by using a Fisher dimension reduction criterion to obtain a dimension reduced energy feature vector;
step S4: and establishing a ring main unit discharge sound signal identification model, and inputting the energy feature vector after the dimension reduction into the ring main unit discharge sound signal identification model to obtain a ring main unit discharge sound signal identification result.
Further, in the step S1, the method further includes:
and filtering the collected ring main unit operation sound signals to remove power frequency and harmonic noise thereof through a band-pass filter.
Further, in the step S2, the method further includes:
step S21: after decomposing the ring main unit operation sound signals to each frequency segment by a wavelet packet energy analysis method, dividing the frequency segment of 0-48khz, and decomposing by a 5-layer wavelet packet to obtain waveform signals of 32 sub-frequency segments, wherein each frequency segment interval is 1.5khz;
step S22: the energy of the decomposed signal on each sub-frequency segment is calculated respectively, so that a wavelet packet node energy feature vector is formed; wherein, the energy E of the decomposed signal on each sub-frequency section of the ring main unit operation sound signal i And the calculation formula of the total energy E of the running sound signal of the ring main unit is as follows;
wherein x is i (n) is a sequence of decomposed signals on the ith sub-frequency band, n is a sequence of ring main unit operation sound signal sample numbers, for example, 200 ring main unit operation sound signal samples, and n=1 represents a first ring main unit operation sound signal sample;
step S23: calculating the energy E of the decomposed signal on each sub-frequency segment i Percent T of (2) i Percentage T i The calculation formula of (2) is as follows:
wherein T is i For decomposing signals in the ith sub-frequency bandPercent of energy;
step S24: discarding frequency segments below 6khz and frequency segments above 18khz, and reserving only 8 sub-frequency segments, so that an energy characteristic vector T of a ring main unit discharge sound signal is obtained as follows:
the energy characteristic vector T of the ring main unit discharging sound signal represents the percentage of the energy of the decomposed signal on different sub-frequency sections.
Further, in the step S3, the method further includes:
step S31: calculating the importance degree of the energy feature vector of the ring main unit discharge sound signal according to Fisher dimension reduction criteria;
wherein, the Fisher dimension reduction criterion is:
wherein r is Fisher The Fisher ratio of the energy characteristic vector of the ring main unit discharge sound signal is calculated by the Fisher dimension reduction criterion; sigma (sigma) between The inter-class variance of the energy characteristic vector is represented, namely the variance of the energy characteristic component mean value of the discharge sound signals of different ring main units; sigma (sigma) within The intra-class variance of the energy characteristic vector is represented, namely, the average value of the variances of the energy characteristic components of the discharge sound signals of the same ring main unit;
wherein k represents the dimension of the energy feature vector; m is m k A mean value representing the kth component of the energy feature vectors of all classes;a mean value representing the kth component of the ith class in the acoustic energy feature vector; omega i A sound feature sequence representing class i; n represents the number of categories of the classified sound feature sequence; n is n i C is a component of the sound feature sequence;
step S32: the dimension reduction method of the Fisher ratio is calculated, the dimension of the original 8-dimensional energy feature vector is reduced to 6 dimensions, and a new energy feature vector formed by 6 frequency segments with larger discrimination is selected as the basis of final identification.
Further, in the step S4, the method further includes: identifying the energy feature vector after the dimension reduction by using a single-classification support vector machine mode so as to identify the type of the ring main unit discharge sound signal;
step S41: the energy feature vector of the ring main unit discharge sound signal after dimension reduction is mapped to a high-dimension feature space through nonlinear mapping to form a closed hypersphere, and the method comprises the following steps:
is provided with R N Nonlinear mapping phi to a high-dimensional feature space χ is such that phi (X i ) E χ, find a super sphere S with radius R sphere center a, let phi (X i ) As far as possible contained in the hypersphere S, the following constraint equation is obtained:
wherein, xi i C is a relaxation coefficient for relaxation variables, and a very small part of training data is allowed to not meet constraint requirements;
in order to simplify the dimension of the established high-dimensional feature space, the high-dimensional feature space is mapped through a kernel function, and the Gaussian kernel function is as follows: k (x) i )＝exp{-γx i T x i - γ is a gaussian kernel width parameter;
step S42: the energy feature vectors of the ring main unit discharge sound signals after dimension reduction are classified according to the positions of the hyperspheres, and the following Lagrange equation is obtained according to the constraint equation of the hyperspheres:
wherein a is i ≥0,β i 0 is Lagrangian operator;
if a is i =0, then c=β i ，ξ i The position of the energy feature vector of the ring main unit discharge sound signal after the dimension reduction is in the super sphere S, which indicates that the energy feature vector of the ring main unit discharge sound signal after the dimension reduction belongs to the category; if 0 is<a i <C, then R 2 +ξ i -||φ(x i -a)|| 2 =0, simultaneously satisfy ζ i Not less than 0, wherein the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction is not in the super sphere S, which indicates that the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction does not belong to the category; denoted as d 2 (x)＝||φ(x i -a)|| 2 For the distance from the energy characteristic vector to the sphere center, judging whether the ring main unit to be classified discharges the sound signal to be the energy characteristic vector of the insulator surface discharge, namely:
f(x)＝sgn(R 2 -||φ(x i -a)|| 2 )＝sgn(R 2 -d 2 (x))
wherein R is the distance from the energy characteristic vector to the sphere center;
step S43: and constructing an OC-SVM identification model, wherein only the energy feature vector of the ring main unit discharging sound signal after the dimension reduction is trained in the training process.
The ring main unit discharge state identification method based on the audio signal has the following advantages: the energy distribution of sound data can be accurately used for judging the insulation state by adopting wavelet packet energy analysis to characterize discharge characteristics; performing dimension reduction on the extracted feature vector by using a Fisher dimension reduction criterion; based on the classification condition of the ring main unit discharging sound, a recognition model for on-line sound monitoring based on a single classification support vector machine is provided in combination with reality, the comprehensive recognition rate is close to hundred percent, and compared with the traditional ring main unit fault monitoring method, the method has better recognition precision and is more rapid and reliable.
Drawings
The accompanying drawings are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate the invention and together with the description serve to explain, without limitation, the invention.
Fig. 1 is a flowchart of a ring main unit discharge state identification method based on audio signals.
Fig. 2 is a schematic diagram of a classification mode of a single classification support vector machine provided by the invention.
FIG. 3 is a bar graph of Fisher ratios provided by the present invention.
Fig. 4 is a flowchart of the OC-SVM algorithm provided by the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to specific implementation, structure, characteristics and effects of the ring main unit discharge state identification method based on audio signals according to the invention by combining the accompanying drawings and the preferred embodiment. It will be apparent that the described embodiments are some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
In this embodiment, an audio signal-based ring main unit discharge state identification method is provided, as shown in fig. 1, where the audio signal-based ring main unit discharge state identification method includes:
step S1: collecting ring main unit operation sound signals in different states, wherein the ring main unit operation sound signals in different states comprise ring main unit discharge sound signals in different types;
it should be noted that, each section of test sound is recorded for 2 seconds, different insulators are recorded for 10 times in different states, and the test sound is decomposed into data of one section of 100ms, and 200 groups of ring main unit operation sound sample data are obtained in each state;
step S2: processing the ring main unit operation sound signal to obtain an energy characteristic vector of a ring main unit discharge sound signal, wherein the energy characteristic vector of the ring main unit discharge sound signal represents the characteristic of a discharge sound signal waveform;
step S3: reducing the dimension of the energy feature vector of the ring main unit discharge sound signal by using a Fisher dimension reduction criterion to obtain a dimension reduced energy feature vector;
the characteristics of the discharge sound signal waveform can be intensively represented, the energy feature vector after dimension reduction has the characteristic of being close to Gaussian distribution, and the characteristics of input sample data in the ring main unit discharge sound signal recognition model training stage are met;
step S4: and establishing a ring main unit discharge sound signal identification model, and inputting the energy feature vector after the dimension reduction into the ring main unit discharge sound signal identification model to obtain a ring main unit discharge sound signal identification result.
It should be understood that the energy feature vectors after the dimension reduction are identified by the ring main unit discharge sound signal identification model to distinguish between different types of discharge sound signals.
Preferably, in the step S1, the method further includes:
and filtering the collected ring main unit operation sound signals to remove power frequency and harmonic noise thereof through a band-pass filter.
Preferably, in the step S2, the method further includes:
step S21: decomposing the ring main unit operation sound signals (200 groups of ring main unit operation sound sample data) onto each frequency segment by a wavelet packet energy analysis method, dividing the frequency segment of 0-48khz, and decomposing by 5 layers of wavelet packets to obtain waveform signals of 32 sub-frequency segments, wherein each frequency segment interval is 1.5khz;
step S22: the energy of the decomposed signal on each sub-frequency segment is calculated respectively, so that a wavelet packet node energy feature vector is formed; wherein, each sub-frequency section of the ring main unit operation sound signalUp-decomposing the energy E of the signal i And the calculation formula of the total energy E of the running sound signal of the ring main unit is as follows;
wherein x is i (n) is a sequence of decomposed signals on the ith sub-frequency band, n is a sequence of ring main unit operation sound signal sample numbers, for example, 200 ring main unit operation sound signal samples, and n=1 represents a first ring main unit operation sound signal sample;
step S23: in order to facilitate analysis and processing of data, the analysis result is more visual, and the obtained energy is normalized according to the following formula, specifically, the energy E of the decomposed signal on each sub-frequency segment is calculated respectively i Percent T of (2) i Percentage T i The calculation formula of (2) is as follows:
wherein T is i The percentage of the decomposed signal energy over the ith sub-frequency bin;
step S24: considering the effect of the band-pass filter, discarding frequency bands below 6khz and frequency bands above 18khz, and only reserving 8 sub-frequency bands, so that the energy characteristic vector T of the ring main unit discharge sound signal is obtained as follows:
the energy characteristic vector T of the ring main unit discharging sound signal represents the percentage of the energy of the decomposed signal on different sub-frequency sections.
It should be noted that, based on wavelet packet analysis, an energy feature vector of the discharge sound signal is obtained, which characterizes the energy of the frequency band of the discharge sound signal. In the stage of the recognition training algorithm, excessive dimensions can reduce the recognition efficiency and accuracy. The energy components of certain frequency ranges cannot be distinguished from other sounds, nor can they characterize the discharge sound signal. In order to further optimize the data, the Fisher criterion is required to evaluate the influence degree of the energy feature vector, so as to obtain the feature component with larger distinguishing degree and form a new frequency band energy feature vector.
Preferably, in the step S3, the method further includes:
step S31: calculating the importance degree of the energy feature vector of the ring main unit discharge sound signal according to Fisher dimension reduction criteria;
wherein, the Fisher dimension reduction criterion is:
wherein r is Fisher The Fisher ratio of the energy characteristic vector of the ring main unit discharge sound signal is calculated by the Fisher dimension reduction criterion; sigma (sigma) between The inter-class variance of the energy characteristic vector is represented, namely the variance of the energy characteristic component mean value of the discharge sound signals of different ring main units; sigma within represents the intra-class variance of the energy feature vector, namely the mean value of the variances of the energy feature components of the discharge sound signals of the same ring main unit;
wherein k represents the dimension of the energy feature vector; m is m k A mean value representing the kth component of the energy feature vectors of all classes;a mean value representing the kth component of the ith class in the acoustic energy feature vector; omega i A sound feature sequence representing class i; n represents the number of categories of the classified sound feature sequence; n is n i C is a component of the sound feature sequence;
the inter-class variance σ of the energy feature component between Characterizing the differentiation between different types of discharge sounds, while the intra-class variance sigma within Reflecting the association degree of the discharge sound signals; for the sound feature vectors of different categories, the intra-category variance and the inter-category variance of the discharge sound signals of different categories need to be calculated to obtain the relation between Fisher criteria of the classification quantities of different frequency bands of the discharge sound signals;
step S32: the dimension reduction method of the Fisher ratio is calculated, the dimension of the original 8-dimensional energy feature vector is reduced to 6 dimensions, and a new energy feature vector formed by 6 frequency segments with larger discrimination is selected as a basis for final identification; obtaining Fisher ratio of different frequency intervals, and eliminating characteristic component with Fser ratio less than 0.1, namely E 7 ,E 8 Two components, reserve E 1 ～E 6 A component; the Fisher ratio results are shown in FIG. 3, and the dimension of the energy eigenvector is reduced by dimension reduction, so that the characteristics of the energy eigenvector can be represented in a concentrated manner.
Preferably, in the step S4, the method further includes: identifying the energy feature vector after the dimension reduction by using a single-classification support vector machine mode so as to identify the type of the ring main unit discharge sound signal;
step S41: the energy feature vector of the ring main unit discharge sound signal after dimension reduction is mapped to a high-dimension feature space through nonlinear mapping to form a closed hypersphere, and the method comprises the following steps:
is provided with R N Nonlinear mapping phi to a high-dimensional feature space χ is such that phi (X i ) E χ, find a super sphere S with radius R sphere center a, let phi (X i ) As far as possible contained in the hypersphere S, the following constraint equation is obtained:
wherein, xi i C is a relaxation coefficient for relaxation variables, and a very small part of training data is allowed to not meet constraint requirements;
in order to simplify the dimension of the established high-dimensional feature space, the high-dimensional feature space is mapped through a kernel function, and the Gaussian kernel function is as follows: k (x) i )＝exp{-γx i T x i - γ is a gaussian kernel width parameter;
step S42: the energy feature vectors of the ring main unit discharge sound signals after dimension reduction are classified according to the positions of the hyperspheres, and the following Lagrange equation is obtained according to the constraint equation of the hyperspheres:
wherein a is i ≥0,β i 0 is Lagrangian operator;
if a is i =0, then c=β i ，ξ i The position of the energy feature vector of the ring main unit discharge sound signal after the dimension reduction is in the super sphere S, which indicates that the energy feature vector of the ring main unit discharge sound signal after the dimension reduction belongs to the category; if 0 is<a i <C, then R 2 +ξ i -||φ(x i -a)|| 2 =0, simultaneously satisfy ζ i Not less than 0, wherein the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction is not in the super sphere S, which indicates that the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction does not belong to the category; denoted as d 2 (x)＝||φ(x i -a)|| 2 For the distance from the energy characteristic vector to the sphere center, judging whether the ring main unit to be classified discharges the sound signal to be the energy characteristic vector of the insulator surface discharge, namely:
f(x)＝sgn(R 2 -||φ(x i -a)|| 2 )＝sgn(R 2 -d 2 (x))
wherein R is the distance from the energy characteristic vector to the sphere center; the classification mode of the single-classification support vector machine is shown in fig. 2;
step S43: and constructing an OC-SVM identification model, wherein only the energy feature vector of the ring main unit discharging sound signal after the dimension reduction is trained in the training process.
It should be noted that the recognition rate of the OC-SVM recognition model mainly depends on the kernel width parameter γ and the relaxation coefficient C, and a higher recognition rate is obtained by adjusting the two parameters. The core width parameter gamma and the relaxation coefficient C are continuously changed by a grid search method, so that the optimal parameter combination is obtained.
Specifically, the grid search method is an exhaustive method for parameters, and specifically comprises the following steps: the possible values of the nuclear parameters and the relaxation parameters are listed for arrangement and combination, and the combination forms a grid; training each obtained possible parameter combination, and evaluating each parameter combination by using a cross-validation method; continuously changing a parameter gamma and a parameter C through grid search, and searching the optimal recognition rate; through continuous loop iteration, the identification effect under the combination condition of different parameters gamma and parameters C can be obtained, the optimal identification parameter combination is selected, the comprehensive identification rate of sound can reach 99.93% through calculation, and the method has practical application feasibility; the OC-SVM algorithm flow is shown in FIG. 4.
According to the ring main unit discharge state identification method based on the audio signals, provided by the invention, the discharge characteristics are characterized by wavelet packet energy analysis, and the insulation state can be accurately judged by the energy distribution of sound data; performing dimension reduction on the extracted feature vector by using a Fisher dimension reduction criterion; based on the classification condition of the ring main unit discharging sound, a recognition model for on-line sound monitoring based on a single classification support vector machine is provided in combination with reality, the comprehensive recognition rate is close to hundred percent, and compared with the traditional ring main unit fault monitoring method, the method has better recognition precision and is more rapid and reliable.
The present invention is not limited to the above-mentioned embodiments, but is intended to be limited to the following embodiments, and any modifications, equivalents and modifications can be made to the above-mentioned embodiments without departing from the scope of the invention.
Claims (4)
1. The ring main unit discharge state identification method based on the audio signal is characterized by comprising the following steps of:
step S1: collecting ring main unit operation sound signals in different states, wherein the ring main unit operation sound signals in different states comprise ring main unit discharge sound signals in different types;
step S2: processing the ring main unit operation sound signal to obtain an energy characteristic vector of a ring main unit discharge sound signal, wherein the energy characteristic vector of the ring main unit discharge sound signal represents the characteristic of a discharge sound signal waveform;
step S3: reducing the dimension of the energy feature vector of the ring main unit discharge sound signal by using a Fisher dimension reduction criterion to obtain a dimension reduced energy feature vector;
step S4: establishing a ring main unit discharge sound signal identification model, and inputting the energy feature vector subjected to the dimension reduction into the ring main unit discharge sound signal identification model to obtain a ring main unit discharge sound signal identification result;
wherein, in the step S4, the method further comprises: identifying the energy feature vector after the dimension reduction by using a single-classification support vector machine mode so as to identify the type of the ring main unit discharge sound signal;
step S41: the energy feature vector of the ring main unit discharge sound signal after dimension reduction is mapped to a high-dimension feature space through nonlinear mapping to form a closed hypersphere, and the method comprises the following steps:
is provided with R N Nonlinear mapping phi to a high-dimensional feature space χ is such that phi (X i ) E chi, seekFind a super sphere S with radius R and sphere center a, let phi (X i ) As far as possible contained in the hypersphere S, the following constraint equation is obtained:
wherein, xi i C is a relaxation coefficient for relaxation variables, and a very small part of training data is allowed to not meet constraint requirements;
in order to simplify the dimension of the established high-dimensional feature space, the high-dimensional feature space is mapped through a kernel function, and the Gaussian kernel function is as follows: k (x) i )＝exp{-γx i T x i - γ is a gaussian kernel width parameter;
step S42: the energy feature vectors of the ring main unit discharge sound signals after dimension reduction are classified according to the positions of the hyperspheres, and the following Lagrange equation is obtained according to the constraint equation of the hyperspheres:
wherein a is i ≥0,β i 0 is Lagrangian operator;
if a is i =0, then c=β i ，ξ i The position of the energy feature vector of the ring main unit discharge sound signal after the dimension reduction is in the super sphere S, which indicates that the energy feature vector of the ring main unit discharge sound signal after the dimension reduction belongs to the category; if 0 is<a i <C, then R 2 +ξ i -||φ(x i -a)|| 2 =0, simultaneously satisfy ζ i Not less than 0, wherein the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction is not in the super sphere S, which indicates that the energy characteristic vector of the ring main unit discharge sound signal after the dimension reduction does not belong to the category; denoted as d 2 (x)＝||φ(x i -a)|| 2 For the distance from the energy characteristic vector to the sphere center, judging whether the ring main unit to be classified discharges the sound signal to be the energy characteristic vector of the insulator surface discharge,namely:
f(x)＝sgn(R 2 -||φ(x i -a)|| 2 )＝sgn(R 2 -d 2 (x))
wherein R is the distance from the energy characteristic vector to the sphere center;
step S43: and constructing an OC-SVM identification model, wherein only the energy feature vector of the ring main unit discharging sound signal after the dimension reduction is trained in the training process.
2. The method for identifying a discharge state of a ring main unit based on an audio signal according to claim 1, wherein in step S1, further comprises:
and filtering the collected ring main unit operation sound signals to remove power frequency and harmonic noise thereof through a band-pass filter.
3. The method for identifying a discharge state of a ring main unit based on an audio signal according to claim 1, wherein in step S2, further comprises:
step S21: after decomposing the ring main unit operation sound signals to each frequency segment by a wavelet packet energy analysis method, dividing the frequency segment of 0-48khz, and decomposing by a 5-layer wavelet packet to obtain waveform signals of 32 sub-frequency segments, wherein each frequency segment interval is 1.5khz;
step S22: the energy of the decomposed signal on each sub-frequency segment is calculated respectively, so that a wavelet packet node energy feature vector is formed; wherein, the energy E of the decomposed signal on each sub-frequency section of the ring main unit operation sound signal i And the calculation formula of the total energy E of the running sound signal of the ring main unit is as follows;
wherein x is i (n) is on the ith sub-frequency bandN is a sequence of ring main unit operation sound signal sample numbers, for example, 200 ring main unit operation sound signal samples, and n=1 represents a first ring main unit operation sound signal sample;
step S23: calculating the energy E of the decomposed signal on each sub-frequency segment i Percent T of (2) i Percentage T i The calculation formula of (2) is as follows:
wherein T is i The percentage of the decomposed signal energy over the ith sub-frequency bin;
step S24: discarding frequency segments below 6khz and frequency segments above 18khz, and reserving only 8 sub-frequency segments, so that an energy characteristic vector T of a ring main unit discharge sound signal is obtained as follows:
the energy characteristic vector T of the ring main unit discharging sound signal represents the percentage of the energy of the decomposed signal on different sub-frequency sections.
4. A method for identifying a discharge state of a ring main unit based on an audio signal according to claim 3, wherein in step S3, the method further comprises:
step S31: calculating the importance degree of the energy feature vector of the ring main unit discharge sound signal according to Fisher dimension reduction criteria;
wherein, the Fisher dimension reduction criterion is:
wherein r is Fisher The Fisher ratio of the energy characteristic vector of the ring main unit discharge sound signal is calculated by the Fisher dimension reduction criterion; sigma (sigma) between Representing energyThe inter-class variance of the feature vector is the variance of the energy feature component mean value of the discharge sound signals of different ring main units; sigma (sigma) within The intra-class variance of the energy characteristic vector is represented, namely, the average value of the variances of the energy characteristic components of the discharge sound signals of the same ring main unit;
wherein k represents the dimension of the energy feature vector; m is m k A mean value representing the kth component of the energy feature vectors of all classes;a mean value representing the kth component of the ith class in the acoustic energy feature vector; omega i A sound feature sequence representing class i; n represents the number of categories of the classified sound feature sequence; n is n i C is a component of the sound feature sequence;
step S32: the dimension reduction method of the Fisher ratio is calculated, the dimension of the original 8-dimensional energy feature vector is reduced to 6 dimensions, and a new energy feature vector formed by 6 frequency segments with larger discrimination is selected as the basis of final identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110868613.7A CN113608082B (en) | 2021-07-30 | 2021-07-30 | Ring main unit discharge state identification method based on audio signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110868613.7A CN113608082B (en) | 2021-07-30 | 2021-07-30 | Ring main unit discharge state identification method based on audio signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113608082A CN113608082A (en) | 2021-11-05 |
| CN113608082B true CN113608082B (en) | 2024-03-22 |
Family
ID=78306118
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110868613.7A Active CN113608082B (en) | 2021-07-30 | 2021-07-30 | Ring main unit discharge state identification method based on audio signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113608082B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102426835A (en) * | 2011-08-30 | 2012-04-25 | 华南理工大学 | Method for identifying local discharge signals of switchboard based on support vector machine model |
| WO2013177981A1 (en) * | 2012-05-28 | 2013-12-05 | 中兴通讯股份有限公司 | Scene recognition method, device and mobile terminal based on ambient sound |
| CN204349328U (en) * | 2015-01-28 | 2015-05-20 | 国家电网公司 | A kind of SF6 gas-insulated ring network cabinet with local discharge on-line monitoring device |
| CN108303624A (en) * | 2018-01-31 | 2018-07-20 | 舒天才 | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis |
| CN108896878A (en) * | 2018-05-10 | 2018-11-27 | 国家电网公司 | A kind of detection method for local discharge based on ultrasound |
-
2021
- 2021-07-30 CN CN202110868613.7A patent/CN113608082B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102426835A (en) * | 2011-08-30 | 2012-04-25 | 华南理工大学 | Method for identifying local discharge signals of switchboard based on support vector machine model |
| WO2013177981A1 (en) * | 2012-05-28 | 2013-12-05 | 中兴通讯股份有限公司 | Scene recognition method, device and mobile terminal based on ambient sound |
| CN204349328U (en) * | 2015-01-28 | 2015-05-20 | 国家电网公司 | A kind of SF6 gas-insulated ring network cabinet with local discharge on-line monitoring device |
| CN108303624A (en) * | 2018-01-31 | 2018-07-20 | 舒天才 | A kind of method for detection of partial discharge of switch cabinet based on voice signal analysis |
| CN108896878A (en) * | 2018-05-10 | 2018-11-27 | 国家电网公司 | A kind of detection method for local discharge based on ultrasound |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113608082A (en) | 2021-11-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103076547B (en) | Method for identifying GIS (Gas Insulated Switchgear) local discharge fault type mode based on support vector machines | |
| CN111814872B (en) | Power equipment environmental noise identification method based on time domain and frequency domain self-similarity | |
| CN111239672B (en) | Machine learning algorithm-based gradient fault prediction method for optical fiber current transformer | |
| CN108802525A (en) | Equipment fault intelligent Forecasting based on small sample | |
| CN109443717B (en) | On-load tap-changer mechanical fault on-line monitoring method | |
| CN114492898A (en) | Product failure prediction method and device, electronic device and storage medium | |
| CN107798283A (en) | A kind of neural network failure multi classifier based on the acyclic figure of decision-directed | |
| Catterson et al. | Identifying harmonic attributes from online partial discharge data | |
| Huang et al. | Feature extraction for gas metal arc welding based on EMD and time–frequency entropy | |
| CN112486137A (en) | Method and system for constructing fault feature library of active power distribution network and fault diagnosis method | |
| CN110794254A (en) | Power distribution network fault prediction method and system based on reinforcement learning | |
| CN113608082B (en) | Ring main unit discharge state identification method based on audio signals | |
| CN117786538A (en) | CsAdaBoost integrated learning algorithm based on cost sensitivity improvement | |
| CN111079647A (en) | Circuit breaker defect identification method | |
| CN111933186B (en) | Method, device and system for fault identification of on-load tap-changer | |
| Dang et al. | Fault diagnosis of power transformer by acoustic signals with deep learning | |
| Xu et al. | A vibration signal anomaly detection method based on frequency component clustering and isolated forest algorithm | |
| CN109635430A (en) | Grid power transmission route transient signal monitoring method and system | |
| CN117666406A (en) | Multi-parameter synchronous dynamic signal acquisition method and system based on edge calculation | |
| CN114609435A (en) | Voltage sag detection and classification identification method | |
| CN109917245B (en) | Ultrasonic detection partial discharge signal mode identification method considering phase difference | |
| CN112729531B (en) | Fault studying and judging method and system for distribution transformer equipment | |
| CN107271024A (en) | A kind of load ratio bridging switch diagnostic method and device | |
| CN114121025A (en) | Voiceprint fault intelligent detection method and device for substation equipment | |
| Luo et al. | Insulation fault identification of vacuum circuit breakers based on improved MFCC and SVM |
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 |