CN117153193B - Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis - Google Patents
Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis Download PDFInfo
- Publication number
- CN117153193B CN117153193B CN202311412959.1A CN202311412959A CN117153193B CN 117153193 B CN117153193 B CN 117153193B CN 202311412959 A CN202311412959 A CN 202311412959A CN 117153193 B CN117153193 B CN 117153193B
- Authority
- CN
- China
- Prior art keywords
- voiceprint
- signal
- normal
- abnormal
- power equipment
- 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
- 238000003745 diagnosis Methods 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 43
- 230000002159 abnormal effect Effects 0.000 claims abstract description 114
- 108010076504 Protein Sorting Signals Proteins 0.000 claims abstract description 67
- 238000012216 screening Methods 0.000 claims abstract description 34
- 230000015556 catabolic process Effects 0.000 claims abstract description 28
- 238000006731 degradation reaction Methods 0.000 claims abstract description 28
- 239000000203 mixture Substances 0.000 claims abstract description 28
- 238000012544 monitoring process Methods 0.000 claims description 50
- 230000005236 sound signal Effects 0.000 claims description 23
- 239000013598 vector Substances 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 15
- 238000004458 analytical method Methods 0.000 description 14
- 238000012549 training Methods 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 7
- 238000010801 machine learning Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000007547 defect Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000005856 abnormality Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000001960 triggered effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000004587 chromatography analysis Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000001931 thermography Methods 0.000 description 2
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000009422 external insulation Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000009421 internal insulation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- 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
-
- 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
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
Abstract
The invention discloses a power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis, which comprises the steps of collecting power equipment voiceprint signals and generating a time domain discrete digital signal sequence according to a time sequence; calculating the sound intensity variation degree and the sound intensity variation degree of the time domain discrete digital signal sequence, and screening out a first-stage abnormal voiceprint signal and a first-stage normal voiceprint signal; diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal; performing fault diagnosis on the second-level abnormal voiceprint signal, and determining a fault type; triggering early warning based on the first-level abnormal voiceprint signal and the fault type; the invention fully utilizes the original voiceprint signal, and can realize accurate and rapid diagnosis of the fault voiceprint of the power equipment.
Description
Technical Field
The invention relates to the technical field of equipment fault diagnosis, in particular to a power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis.
Background
The novel power system is a key for supporting energy green low-carbon transformation, safety and high efficiency are important preconditions for constructing the novel power system, the power equipment is a material foundation of the novel power system, and the effective state monitoring means are main technical measures for ensuring safe and stable operation of the power equipment.
The traditional power equipment state monitoring technology comprises oil chromatography monitoring, partial discharge monitoring, infrared thermal imaging monitoring, vibration monitoring and the like, and the monitoring means have advantages and disadvantages, so that the oil chromatography monitoring can effectively discover the defects of oil media in the power equipment, but the sensitivity of the reaction is lower; partial discharge monitoring can effectively find out internal and external insulation discharge defects of power equipment, but is easy to be interfered by the outside to influence the monitoring accuracy; the infrared thermal imaging monitoring can effectively find out the voltage current type heating defect of the power equipment, but the monitoring effect has a larger relation with the level of operators; vibration monitoring can effectively find mechanical defects of electric equipment, but a contact monitoring mode is needed, and the monitoring range is limited.
The voiceprint monitoring technical means is used as a novel detection mode, the types of defects such as machinery, electromagnetism and the like inside and outside the power equipment can be effectively found, various information processing technologies are utilized to analyze voiceprint signals, for example, a time-frequency method is utilized to analyze key characteristics such as main frequency, amplitude and the like of the voiceprint signals, the similar voiceprint signals can be trained, learned and identified through a machine learning model, so that the voiceprint characteristics of the power equipment can be identified quickly, meanwhile, the voiceprint identification can be implemented without power failure, non-contact monitoring is realized, and safety and reliability in monitoring are guaranteed, for example, the fault detection method of the power equipment based on the voiceprint identification is described in patent application document with publication number CN 115376526A.
The existing voiceprint recognition method can realize the diagnosis of the faults of the power equipment, but the voiceprint signals are not fully applied, the traditional time-frequency method only uses signals of a time domain and a frequency domain, the information characteristics of higher dimensionality are lost, the calculation and processing time is longer, the dependence on the physical operation characteristics of the power equipment is stronger, the type of the recognized faults is limited, the recognition accuracy of a machine learning model is lower under the condition of lacking a large number of training samples, and the interpretation of the recognition process and the conclusion is poorer.
Disclosure of Invention
The invention aims to solve the technical problem of how to fully utilize the original voiceprint signal to realize accurate and rapid diagnosis of the fault voiceprint of the power equipment.
The invention solves the technical problems by the following technical means:
the invention provides a power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis, which comprises the following steps:
collecting voiceprint signals of the power equipment, and generating a time domain discrete digital signal sequence according to a time sequence;
calculating the sound intensity variation degree and the sound intensity variation degree of the time domain discrete digital signal sequence, and screening out a first-stage abnormal voiceprint signal and a first-stage normal voiceprint signal;
Diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
performing fault diagnosis on the second-level abnormal voiceprint signal, and determining a fault type;
and triggering early warning based on the first-level abnormal voiceprint signal and the fault type.
Further, the collecting the voiceprint signal of the electric equipment and generating a time domain discrete digital signal sequence according to a time sequence includes:
recording the sampling rate f and the monitoring period T of the voiceprint signals to obtain the total number N of discrete digital signals in one monitoring period;
and numbering the discrete digital signals in one monitoring period, and generating the time domain discrete digital signal sequence according to the time sequence.
Further, the calculating the sound intensity variation and the sound frequency variation of the time domain discrete digital signal sequence, and screening out the first-stage abnormal voiceprint signal and the first-stage normal voiceprint signal includes:
performing sound intensity variation degree identification on the time domain discrete digital signal sequence, and screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals;
performing discrete Fourier transform on the time domain discrete digital signal sequence to obtain a frequency domain discrete digital signal sequence;
Carrying out audio frequency variation degree identification on the frequency domain discrete digital signal sequence, and screening out an abnormal audio voiceprint signal and an normal audio voiceprint signal;
and taking the sound intensity abnormal voiceprint signal and the audio frequency abnormal voiceprint signal as the first-stage abnormal voiceprint signal, and taking the sound intensity normal voiceprint signal and the audio frequency normal voiceprint signal as the first-stage normal voiceprint signal.
Further, the abnormal sound intensity voiceprint signal includes short overload, long overload, sudden trip short circuit and overload fault, the sound intensity variation degree identification is performed on the time domain discrete digital signal sequence, and the abnormal sound intensity voiceprint signal and the normal sound intensity voiceprint signal are screened out, including:
calculating the sum of sound intensities of the voiceprint signals according to the time domain discrete digital signal sequence;
according to the sum of the sound intensities of the sound signals of the current power equipment and the sound intensities of the sound signals of the adjacent power equipment, determining that the sound signals of the current power equipment are sound intensity normal sound signals or short-time overload or long-time overload or sudden tripping short circuit;
in a power grid structure with a symmetrical A, B, C three-phase structure, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print or an overload fault according to the average value of voice print signal voice print sums of the three-phase power equipment.
Further, when (L N+1 －L N ）＞0，And is also provided withWhen the variation range is within a first threshold, determining that short-time overload occurs in the current power equipment;
when (L) N+1 －L N ）＞0，And->When the current power equipment grows, determining overload;
when (L) N+1 －L N ) Not equal to 0, andwhen determining that the current power equipment has sudden trip short circuit, wherein L N Is the sum of the sound intensities of the voiceprint signals of the current power equipment, L N+1 And L N+2 Respectively summing the sound intensities of the sound signals of the adjacent devices of the current power device;
and when the deviation degree of a certain phase is larger than a second threshold value, determining that the phase has overload faults.
Further, the abnormal audio voiceprint signal includes a harmonic working condition and a dc magnetic bias working condition, the identifying the audio variation degree of the frequency domain discrete digital signal sequence, and screening out the abnormal audio voiceprint signal and the normal audio voiceprint signal includes:
and determining that the voiceprint signal of the current power equipment is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition according to the amplitude and the duty ratio of frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence.
Further, the determining, according to the amplitude and the duty ratio of the frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence, that the voiceprint signal of the current power device is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition includes:
comparing three in two adjacent monitoring periodsSubharmonic Y 150 Judging whether the current power equipment is in a harmonic working condition or not;
when (when)When the current power equipment is determined to be in the DC magnetic bias working condition, G M For the maximum frequency component duty ratio in the mth monitoring period, G M+1 Is the maximum frequency component duty cycle in the m+1th monitoring period.
Further, the diagnosing the primary normal voiceprint signal by using a gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal, including:
learning m periods of the primary normal voiceprint signals to form a regression prediction model;
predicting the first-level normal voiceprint signal which is not learned by using the regression prediction model, and predicting the result S Y Comparing the residual K with an actual voiceprint signal Ss to form a residual K;
continuously correcting model parameters by using the residual error K, and determining model parameters of a Gaussian mixture degradation prediction algorithm;
And diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm after determining model parameters, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal.
Further, the performing fault diagnosis on the second-level abnormal voiceprint signal, and determining a fault type includes:
processing the secondary abnormal voiceprint signal by using a pre-trained neural network algorithm model to obtain a feature vector corresponding to the secondary abnormal voiceprint signal;
and comparing the feature vector corresponding to the second-level abnormal voiceprint signal with the feature vector of various faults to determine the fault type.
In addition, the invention also provides a power equipment fault voiceprint recognition system integrating physical characteristics and data diagnosis, which comprises:
the signal acquisition module is used for acquiring voiceprint signals of the power equipment and generating a time domain discrete digital signal sequence according to a time sequence;
the first-stage identification module is used for calculating the sound intensity variation and the sound frequency variation of the time domain discrete digital signal sequence and screening out first-stage abnormal voiceprint signals and first-stage normal voiceprint signals;
the secondary identification module is used for diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
The third-level identification module is used for carrying out fault diagnosis on the second-level abnormal voiceprint signals and determining fault types;
and the early warning module is used for triggering early warning based on the first-level abnormal voiceprint signal and the fault type.
The invention has the advantages that:
(1) According to the invention, three-level diagnosis links are utilized to cooperatively analyze the voiceprint signals of the power equipment, the collected voiceprint signals are firstly subjected to primary physical characteristic recognition, and early warning is triggered according to the recognized primary abnormal voiceprint signals, the primary normal voiceprint signals enter the secondary rapid abnormal recognition, the secondary rapid abnormal recognition adopts a Gaussian mixture degradation prediction algorithm for judging only two states of normal and abnormal, and the obtained secondary normal voiceprint signals do not repeatedly enter the three-level fault recognition links, namely a large number of normal voiceprint signals of the power equipment are removed, so that the algorithm efficiency is improved, and the calculation task of a processor is greatly reduced; according to the invention, through cooperation of three-level diagnosis links, physical characteristics and data diagnosis are fused, so that the original voiceprint signal is effectively utilized to the greatest extent, and accurate and rapid diagnosis of the fault voiceprint of the power equipment is realized.
(2) A limited number of signal analysis algorithm models are arranged in the primary physical characteristic recognition process, the fastest and most accurate fault recognition type is reserved, the problems of complex time-frequency domain analysis and calculation and long processing time are solved, and the practical application space of the algorithm is improved.
(3) The machine learning algorithm of various common typical power equipment faults is provided, and the machine learning algorithm is prefabricated in a fault diagnosis link, so that the problem that the recognition accuracy of a machine learning model is low under the condition of lacking a large number of training samples is solved, meanwhile, the interpretability of the recognition process is improved, and the reliability of recognition conclusion is improved.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a schematic flow chart of a method for identifying a fault voiceprint of an electrical device by combining physical characteristics and data diagnosis according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating fault voiceprint recognition and diagnosis of an electrical device according to an embodiment of the present invention;
FIG. 3 is a schematic block diagram of fault voiceprint recognition of an electrical device according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a primary physical characteristic identification process according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a two-level fast anomaly identification process according to an embodiment of the present invention;
FIG. 6 is a schematic diagram of a three-level fault identification process according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a power equipment fault voiceprint recognition system integrating physical characteristics and data diagnosis according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions in the embodiments of the present invention will be clearly and completely described in the following in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. 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 be within the scope of the invention.
As shown in fig. 1, a first embodiment of the present invention discloses a method for identifying a fault voiceprint of an electrical device by fusing physical characteristics and data diagnosis, the method comprising the steps of:
s10, collecting voiceprint signals of the power equipment, and generating a time domain discrete digital signal sequence according to a time sequence;
it should be noted that, in this embodiment, a piezoelectric or electret microphone sensor may be used to collect and monitor sound waves on the surface of the electrical device or in a certain spatial range.
The analog acoustic wave signal transmitted in the air medium collected by the microphone sensor is subjected to digital signal conversion and discretization processing, so that a time domain discrete digital signal sequence can be obtained.
It should be understood that the power equipment according to the present embodiment includes, but is not limited to, transformers, reactors, combined electric appliances, and the like.
S20, calculating the sound intensity variation degree and the sound frequency variation degree of the time domain discrete digital signal sequence, and screening out a first-stage abnormal voiceprint signal and a first-stage normal voiceprint signal;
in this embodiment, the time domain analysis and the frequency domain analysis are performed on the time domain discrete digital signal sequence, so as to identify the sound intensity variation and the sound frequency variation in the primary physical characteristic of the power equipment, and screen out the fault voiceprint signal and the normal voiceprint signal.
S30, diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
the gaussian mixture degradation prediction algorithm used in the embodiment is a gaussian mixture degradation prediction algorithm with well-determined model parameters, and is used for rapidly diagnosing a first-stage normal voiceprint signal, diagnosing a second-stage normal voiceprint signal and a second-stage abnormal voiceprint signal, transmitting the second-stage abnormal voiceprint signal to a next fault identification process, and performing three-stage fault identification.
S40, performing fault diagnosis on the second-level abnormal voiceprint signal, and determining a fault type;
s50, triggering early warning based on the first-level abnormal voiceprint signal and the fault type.
It should be noted that, in this embodiment, the three-stage diagnosis link is utilized to cooperatively analyze the voiceprint signal of the electrical equipment, the collected voiceprint signal firstly carries out the first-stage physical characteristic recognition, the recognized first-stage normal voiceprint signal is sent to the second-stage rapid abnormal recognition, the second-stage rapid abnormal recognition adopts the gaussian mixture degradation prediction algorithm to rapidly recognize the second-stage normal and abnormal voiceprint signals, only the second-stage abnormal voiceprint signal is sent to the three-stage fault recognition process, and the second-stage normal voiceprint signal does not repeatedly enter the three-stage fault recognition link, thereby eliminating a large number of normal voiceprint signals of the electrical equipment, greatly improving the efficiency of the algorithm and greatly reducing the calculation task of the processor; and through three-level diagnosis links, physical characteristics and data diagnosis are fused, so that the original voiceprint signal is effectively utilized to the greatest extent, and the accurate and rapid diagnosis of the fault voiceprint of the power equipment is realized.
As shown in fig. 2 to 3, in this embodiment, the detected first-level abnormal voiceprint signal, that is, the detected fault is pre-warned, and the detected second-level abnormal voiceprint signal is not pre-warned, so that the second-level abnormal voiceprint signal is further sent to the third-level fault detection process to perform fault type diagnosis.
In one embodiment, the step S10: collecting voiceprint signals of the power equipment, and generating a time domain discrete digital signal sequence according to a time sequence, wherein the method comprises the following steps of:
s11, recording the sampling rate f and the monitoring period T of the voiceprint signals to obtain the total number N of discrete digital signals in one monitoring period;
it should be noted that, in this embodiment, the monitoring period of the voiceprint signal of the electrical equipment is defined as T, where T represents the duration of each monitoring period; defining the sampling rate as f, wherein f represents the number of voiceprint signals of the power equipment which can be acquired per second; n is defined as the number of power device voiceprint signal time domains, n=t×f.
S12, numbering the discrete digital signals in one monitoring period, and generating the time domain discrete digital signal sequence according to a time sequence.
It should be noted that, the discrete digital signals in one monitoring period are numbered, and a discrete digital signal sequence Y is generated according to the time sequence N ，Wherein, is less than or equal to 1i≤N，Y N The sound intensity of the electrical device is represented on a physical characteristic.
In one embodiment, the step S20: calculating the sound intensity variation degree and the sound frequency variation degree of the time domain discrete digital signal sequence, screening out a first-stage abnormal voiceprint signal and a first-stage normal voiceprint signal, and specifically comprising the following steps of:
S21, carrying out sound intensity variation degree identification on the time domain discrete digital signal sequence, and screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals;
in particular, developing Y for a generated time domain discrete digital signal sequence N Calculating the sound intensity variation degree, L N The voice-print signal voice-print sum is defined as the sum of voice-print signal voice-print in a specified monitoring period, and voice-print signal time-domain analysis algorithm is utilized to identify the voice-print change degree in the primary physical characteristics of the electric equipment.
S22, performing discrete Fourier transform on the time domain discrete digital signal sequence to obtain a frequency domain discrete digital signal sequence;
s23, carrying out audio frequency variation degree identification on the frequency domain discrete digital signal sequence, and screening out an abnormal sound track signal and an normal sound track signal;
specifically, the power equipment voiceprint time domain discrete digital signal sequence is converted into a frequency domain discrete digital signal sequence through discrete Fourier transform (Discrete Fourier Transform, DFT), and the voiceprint signal frequency domain analysis algorithm is utilized to identify the audio frequency variation degree in the primary physical characteristics of the power equipment.
S24, taking the sound intensity abnormal voiceprint signal and the audio frequency abnormal voiceprint signal as the first-stage abnormal voiceprint signal, and taking the sound intensity normal voiceprint signal and the audio frequency normal voiceprint signal as the first-stage normal voiceprint signal.
The time domain analysis algorithm in the signal processing technology is utilized, so that faults reflecting the time domain type of the power equipment can be effectively identified, and typical faults such as overload and sudden short-circuit tripping of the power equipment can be identified through analysis corresponding to physical characteristics; by utilizing a frequency domain analysis algorithm in a signal processing technology, faults reflecting the frequency domain type of the power equipment can be effectively identified, and typical abnormal states such as harmonic working conditions, direct current magnetic bias and the like of the power equipment are identified through analysis corresponding to physical characteristics.
In an embodiment, the sound intensity abnormal voiceprint signal includes short overload, long overload, sudden trip short circuit and overload fault, and the step S21: the sound intensity variation degree identification is carried out on the time domain discrete digital signal sequence, and abnormal sound intensity sound track signals and normal sound intensity sound track signals are screened out, and the method comprises the following steps:
s211, calculating the sum of sound intensities of the voiceprint signals according to the time domain discrete digital signal sequence;
specifically, L N The method is defined as the sum of the sound intensities of the voiceprint signals of the power equipment in a specified monitoring period:
wherein, the content is less than or equal to 1i≤N。
S212, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print signal, short overload, long overload or sudden tripping short circuit according to the sum of voice print signal voice print signals of the current power equipment and the sum of voice print signal voice print signals of adjacent power equipment;
Specifically, the sum L of sound intensities of sound signals of sound-stripe signals of three adjacent power equipment is calculated N 、L N+1 、L N+2 And judging whether the power equipment is overloaded or not.
It should be understood that L N+1 、L N+2 And L N Similarly, the details are not repeated here。
S213, in the power grid structure with the symmetrical A, B, C three-phase structure, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print or an overload fault according to the average value of the voice print signal voice print sums of the three-phase power equipment.
The embodiment can also utilize the symmetry of three phases of the power grid A, B, C to identify typical faults such as overload of single-phase power equipment, sudden short-circuit tripping short-circuit and the like.
In one embodiment, several cases of intensity change recognition are:
(1) When (L) N+1 －L N ）＞0，And->When the variation range is within a first threshold value, determining that short-time overload occurs in the current power equipment, wherein the value range of the first threshold value is 0-5%;
(2) When (L) N+1 －L N ）＞0，And is also provided withWhen the current power equipment grows, determining overload;
(3) When (L) N+1 －L N ) Not equal to 0, andwhen determining that the current power equipment has sudden trip short circuit, wherein L N Is the sum of the sound intensities of the voiceprint signals of the current power equipment, L N+1 And L N+2 Respectively summing the sound intensities of the sound signals of the adjacent devices of the current power device;
(4) And when the deviation degree between the sum of the sound intensities of the sound signals of each phase and the average value of the sound intensities of the sound signals of the three-phase power equipment is respectively calculated, determining that the phase has overload faults when the deviation degree of a certain phase is larger than a second threshold value, wherein the value range of the second threshold value is larger than or equal to 10%.
Specifically, in this embodiment, the sum of the sound intensities of the voiceprint signals of the power device A, B, C in the same monitoring period is calculated to be L AN 、L BN 、L CN . Definition X N The average value of the sum of the sound intensities of the voiceprint signals of the three-phase power equipment is as follows:
。
then calculate each phase L separately AN 、L BN 、L CN And X is N Is a deviation P:
wherein,three phases A, B, C can be represented, and when a phase deviates by more than a second threshold (which can be 10%), then the phase is overloaded.
In one embodiment, the audio abnormal voiceprint signal includes a harmonic condition and a dc bias condition, and the step S23: the frequency domain discrete digital signal sequence is subjected to audio frequency variation degree identification, and an abnormal audio voiceprint signal and an normal audio voiceprint signal are screened out, which specifically comprises the following steps:
and determining that the voiceprint signal of the current power equipment is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition according to the amplitude and the duty ratio of frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence.
Specifically, the third harmonic in the power system is the worst, the larger the amplitude of the third harmonic is, the larger the influence on relay protection is, the larger the impact on equipment is, and the third harmonic Y in two adjacent monitoring periods is compared 150 Can judge whether the power equipment is in a harmonic working condition, in particular to judge the third harmonic Y in one monitoring period 150 Is the size of (a) when three harmonics are generatedAnd when the wave content is greater than 1.6%, judging that the current power equipment is in a harmonic working condition.
Transformers in electrical equipment are susceptible to direct current magnetic bias, and when the transformers generate direct current magnetic bias, the transformers can be identified through distribution changes of frequency components. The specific implementation mode is as follows:
maximum frequency component G in mth monitoring period M The ratio of the components is as follows:，/>is the maximum frequency component within the m period;
maximum frequency component G in the (m+1) -th monitoring period M+1 The ratio of the components is as follows:，/>is the maximum frequency component within the m+1 period;
when (when)When the maximum frequency component is obviously changed, the power equipment such as a transformer and the like can be judged to have direct-current magnetic bias.
It should be noted that, in this embodiment, a limited number of signal analysis algorithm models are set in the primary physical characteristic recognition process, so as to ensure the recognition speed and accuracy to the greatest extent, and the signal analysis algorithm with a relatively complex calculation formula or an ambiguous recognition conclusion, such as the characteristics of vibration entropy, total harmonic distortion rate, and the like, is not adopted, so that the accuracy of recognizing the fault type is preferentially ensured. And by arranging a voiceprint signal data separator between the primary physical characteristic recognition and the secondary quick recognition links, the recognized fault voiceprint signals are marked, the recognition links of the secondary and below are not repeated, the non-screened abnormal signals are taken as primary normal voiceprint signals, and the primary normal voiceprint signals directly enter the secondary quick recognition links, as shown in fig. 4, so that the recognition efficiency is improved.
In one embodiment, the step S30: the method comprises the steps of diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal, and comprises the following steps:
s31, learning m periods of the primary normal voiceprint signals to form a regression prediction model;
s32, predicting the first-level normal voiceprint signal which is not learned by utilizing the regression prediction model, and predicting the result S Y Comparing the residual K with an actual voiceprint signal Ss to form a residual K;
s33, continuously correcting model parameters by utilizing residual errors K, and determining model parameters of a Gaussian mixture degradation prediction algorithm;
specifically, as shown in fig. 5, in this embodiment, a power equipment voiceprint signal that is not selected in the primary physical characteristic identification is obtained, learning is performed on the power equipment voiceprint signal, that is, the primary normal voiceprint signal, for a certain period number, for example, m periods, and after learning is completed, a regression prediction model is formed. Machine prediction is carried out on the power equipment voiceprint signals which are not learned by using a regression prediction model, and a prediction result S is obtained Y Comparing with the actual voiceprint signal Ss to form a residual K: k=s Y －Ss。
In this embodiment, m cycles of learning are performed on the primary normal voiceprint signal to form an initial prediction model, the initial prediction model is not necessarily accurate, continuous training iteration is required, model parameters are continuously corrected by using residual error K, after a certain number of iterations, the model parameters are determined, and a mature model which is finally formed is a model of the gaussian mixture degradation prediction algorithm.
S34, diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm after determining model parameters, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal.
It should be noted that, the power equipment voiceprint signal is diagnosed by using the gaussian mixture degradation prediction algorithm after the parameters are determined, because a large number of normal voiceprint samples are used in the model training process, the residual value output by the gaussian mixture degradation prediction algorithm to the normal state of the power equipment is smaller and does not reach the set alarm threshold, but when the power equipment is abnormal or fails, the voiceprint signal changes, the residual value output by the gaussian mixture degradation prediction algorithm becomes larger, and when the set alarm threshold is reached, an alarm is triggered.
Further, the residual alarm threshold can be dynamically adjusted, and the residual thresholds of different power equipment can be subjected to self-defining setting so as to meet the requirements of more power equipment voiceprint monitoring applications.
According to the embodiment, a Gaussian mixture degradation prediction algorithm for judging only two states of normal and abnormal is adopted, and the voice print signal data separator is arranged between the primary identification link and the secondary identification link to mark the identified normal voice print signal, so that the three-stage identification link is not repeated, a large number of normal voice print signals of the power equipment can be filtered rapidly, the analysis workload of the rear end is reduced, and the identification efficiency is improved.
In one embodiment, the step S40: performing fault diagnosis on the second-level abnormal voiceprint signal to determine the fault type, wherein the method specifically comprises the following steps of:
s41, processing the secondary abnormal voiceprint signal by using a pre-trained neural network algorithm model to obtain a feature vector corresponding to the secondary abnormal voiceprint signal;
s42, comparing the feature vector corresponding to the second-level abnormal voiceprint signal with the feature vector of various faults to determine the fault type.
It should be noted that, as shown in fig. 6, the neural network algorithm model used in the three-level fault recognition process of the present embodiment is obtained by training in advance, by obtaining a certain number of fault or abnormal voiceprint samples in the real environment, then establishing the neural network algorithm model, obtaining different network structures through training and optimization, and finally the corresponding neural network algorithm model can diagnose the specific fault or abnormal voiceprint samples. Taking the example of identifying a "partial discharge" fault, the training process is:
(1) The partial discharge voiceprint signal is subjected to dimension reduction treatment and divided into N different sub-bands GF by taking frequency as dimension i Wherein i is equal to or greater than 1 and N is equal to or greater than 1, each sub-channel does not cross each other, and the sub-channels together form a complete frequency band.
(2) The 'partial discharge' voiceprint spectrum signals in each sub-band are subjected to neural network analysis, and the input quantity is [ x y x z ]]Performing convolution operation training on the dimensional voiceprint frequency spectrum to obtain a feature vector GX of the sub-band i The feature vector is then fed into a "partial discharge" classifier to determine whether the unknown abnormal voiceprint signal belongs to "partial discharge".
(3) Feature vector GX obtained for all subbands i Combining to finally obtain a feature vector GX of partial discharge, GX= [ GX ] 1 GX 2 ...GX i... GX N ]。
It should be noted that, after each abnormal voiceprint signal enters the neural network, a feature vector is obtained, and when the feature vector is most matched with the trained feature vector of a specific fault, the early warning of the specific fault is triggered.
The training process of the machine learning algorithm of 15 common typical power equipment faults in table 1 is similar to that described above, and the problem of low recognition accuracy of the machine learning model under the condition of lacking a large number of training samples is solved by training neural network models corresponding to different fault diagnoses and prefabricating the neural network models in a diagnosis link, meanwhile, the interpretation of the recognition process is increased, and the reliability of recognition conclusion is improved.
TABLE 1 15 common faults
Further, when the voiceprint signal identified as the second-level abnormality enters a third-level fault diagnosis link, similarity matching of feature vectors is carried out between the voiceprint signal and 15 types of typical faults and the abnormality, and the fault is the highest in matching degree.
In addition, when the matching degree is lower than 60%, the abnormality is marked as abnormal, specific fault classification processing is not performed, and the manual intervention auxiliary judgment of fault types is carried out.
In addition, as shown in fig. 7, a second embodiment of the present invention discloses a power equipment failure voiceprint recognition system integrating physical characteristics and data diagnosis, the system comprising:
the signal acquisition module 10 is used for acquiring voiceprint signals of the power equipment and generating a time domain discrete digital signal sequence according to a time sequence;
the primary identification module 20 is used for calculating the sound intensity variation and the sound frequency variation of the time domain discrete digital signal sequence and screening out primary abnormal voiceprint signals and primary normal voiceprint signals;
the secondary identifying module 30 is configured to diagnose the primary normal voiceprint signal by using a gaussian mixture degradation prediction algorithm, and screen a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
The third-stage identification module 40 is configured to perform fault diagnosis on the second-stage abnormal voiceprint signal, and determine a fault type;
and the early warning module 50 is used for triggering early warning based on the first-level abnormal voiceprint signal and the fault type.
According to the embodiment, the three-level diagnosis link is utilized to cooperatively analyze the voiceprint signals of the electric equipment, the collected voiceprint signals are firstly subjected to primary physical characteristic identification, the identified primary normal voiceprint signals enter secondary quick abnormal identification, the secondary quick abnormal identification adopts a Gaussian mixture degradation prediction algorithm for judging only two states of normal and abnormal, a large number of normal voiceprint signals of the electric equipment are eliminated, namely, the secondary normal voiceprint signals are not repeatedly entered into the three-level fault identification link, the algorithm efficiency is improved, and the calculation task of a processor is greatly reduced; and by means of three-level diagnosis links, physical characteristics and data diagnosis are fused, the maximum effective utilization of the original voiceprint signals is ensured, and the accurate and rapid diagnosis of the fault voiceprint of the power equipment is realized.
In one embodiment, the signal acquisition module 10 is specifically configured to:
recording the sampling rate f and the monitoring period T of the voiceprint signals to obtain the total number N of discrete digital signals in one monitoring period;
And numbering the discrete digital signals in one monitoring period, and generating the time domain discrete digital signal sequence according to the time sequence.
In one embodiment, the primary identification module 20 specifically includes:
the sound intensity recognition unit is used for recognizing the sound intensity variation degree of the time domain discrete digital signal sequence and screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals;
the time-frequency domain conversion unit is used for carrying out discrete Fourier transform on the time domain discrete digital signal sequence to obtain a frequency domain discrete digital signal sequence;
the audio frequency identification unit is used for carrying out audio frequency variation identification on the frequency domain discrete digital signal sequence and screening out an abnormal audio voiceprint signal and an normal audio voiceprint signal;
and the primary voiceprint signal determining unit is used for taking the sound intensity abnormal voiceprint signal and the audio frequency abnormal voiceprint signal as the primary abnormal voiceprint signal and taking the sound intensity normal voiceprint signal and the audio frequency normal voiceprint signal as the primary normal voiceprint signal.
In an embodiment, the sound intensity recognition unit is specifically configured to:
calculating the sum of sound intensities of the voiceprint signals according to the time domain discrete digital signal sequence;
According to the sum of the sound intensities of the sound signals of the current power equipment and the sound intensities of the sound signals of the adjacent power equipment, determining that the sound signals of the current power equipment are sound intensity normal sound signals or short-time overload or long-time overload or sudden tripping short circuit;
in a power grid structure with a symmetrical A, B, C three-phase structure, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print or an overload fault according to the average value of voice print signal voice print sums of the three-phase power equipment.
Specifically, when (L N+1 －L N ）＞0，And->When the variation range is within a first threshold, determining that short-time overload occurs in the current power equipment;
when (L) N+1 －L N ）＞0，And->When the current power equipment grows, determining overload;
when (L) N+1 －L N ) Not equal to 0, andwhen determining that the current power equipment has sudden trip short circuit, wherein L N Is the sum of the sound intensities of the voiceprint signals of the current power equipment, L N+1 And L N+2 Respectively summing the sound intensities of the sound signals of the adjacent devices of the current power device;
and when the deviation degree of a certain phase is larger than a second threshold value, determining that the phase has overload faults.
In an embodiment, the audio frequency identification unit is specifically configured to:
and determining that the voiceprint signal of the current power equipment is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition according to the amplitude and the duty ratio of frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence.
Further, the determining, according to the amplitude and the duty ratio of the frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence, that the voiceprint signal of the current power device is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition includes:
comparing third harmonic Y in adjacent two monitoring periods 150 Judging whether the current power equipment is in a harmonic working condition or not;
when (when)When the current power equipment is determined to be in the DC magnetic bias working condition, G M For the maximum frequency component duty ratio in the mth monitoring period, G M+1 Is the maximum frequency component duty cycle in the m+1th monitoring period.
In one embodiment, the secondary identification module 30 specifically includes:
the learning unit is used for learning the primary normal voiceprint signal for m periods to form a regression prediction model;
a residual calculation unit for predicting the first-order normal voiceprint signal without learning by using the regression prediction model and predicting the result S Y Comparing the residual K with an actual voiceprint signal Ss to form a residual K;
the parameter optimization unit is used for continuously correcting the model parameters by utilizing the residual error K and determining the model parameters of the Gaussian mixture degradation prediction algorithm;
the second-level voiceprint signal determining unit is used for diagnosing the first-level normal voiceprint signal by utilizing a Gaussian mixture degradation prediction algorithm after model parameters are determined, and the second-level abnormal voiceprint signal and the second-level normal voiceprint signal are screened out from the first-level normal voiceprint signal.
In one embodiment, the three-level identification module 40 is specifically configured to:
processing the secondary abnormal voiceprint signal by using a pre-trained neural network algorithm model to obtain a feature vector corresponding to the secondary abnormal voiceprint signal;
and comparing the feature vector corresponding to the second-level abnormal voiceprint signal with the feature vector of various faults to determine the fault type.
It should be noted that, other embodiments or implementation methods of the power equipment fault voiceprint recognition system integrating physical characteristics and data diagnosis according to the present invention may refer to the above method embodiments, and are not repeated here.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.
Claims (8)
1. A method for identifying a fault voiceprint of an electrical device by combining physical characteristics and data diagnosis, the method comprising:
collecting voiceprint signals of the power equipment, and generating a time domain discrete digital signal sequence according to a time sequence;
calculating the sound intensity variation degree and the sound variation degree of the time domain discrete digital signal sequence, screening out a first-stage abnormal voiceprint signal and a first-stage normal voiceprint signal, and comprising the following steps:
The method comprises the steps of carrying out sound intensity variation identification on the time domain discrete digital signal sequence, screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals, wherein the sound intensity abnormal sound track signals comprise short-time overload, long-time overload, sudden tripping short circuit and overload faults, and the method comprises the steps of carrying out sound intensity variation identification on the time domain discrete digital signal sequence, screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals, and comprising the following steps:
calculating the sum of sound intensities of the voiceprint signals according to the time domain discrete digital signal sequence;
according to the sum of the sound intensities of the sound signals of the current power equipment and the sound intensities of the sound signals of the adjacent power equipment, determining that the sound signals of the current power equipment are sound intensity normal sound signals or short-time overload or long-time overload or sudden tripping short circuit;
in a power grid structure with a symmetrical A, B, C three-phase structure, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print intensity or an overload fault according to the average value of voice print signal voice print sums of the three-phase power equipment;
performing discrete Fourier transform on the time domain discrete digital signal sequence to obtain a frequency domain discrete digital signal sequence;
Carrying out audio frequency variation degree identification on the frequency domain discrete digital signal sequence, and screening out an abnormal audio voiceprint signal and an normal audio voiceprint signal;
taking the sound intensity abnormal voiceprint signal and the audio frequency abnormal voiceprint signal as the first-stage abnormal voiceprint signal, and taking the sound intensity normal voiceprint signal and the audio frequency normal voiceprint signal as the first-stage normal voiceprint signal;
diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
performing fault diagnosis on the second-level abnormal voiceprint signal, and determining a fault type;
and triggering early warning based on the first-level abnormal voiceprint signal and the fault type.
2. The method for identifying a power equipment fault voiceprint by fusing physical characteristics and data diagnosis according to claim 1, wherein the steps of collecting power equipment voiceprint signals and generating a time domain discrete digital signal sequence in time sequence comprise:
recording the sampling rate f and the monitoring period T of the voiceprint signals to obtain the total number N of discrete digital signals in one monitoring period;
And numbering the discrete digital signals in one monitoring period, and generating the time domain discrete digital signal sequence according to the time sequence.
3. The method for identifying a power equipment failure voiceprint by fusing physical characteristics and data diagnosis as recited in claim 1, wherein when (L N+1 －L N ）＞0，And->When the variation range is within a first threshold, determining that short-time overload occurs in the current power equipment;
when (L) N+1 －L N ）＞0，And->When the current power equipment grows, determining overload;
when (L) N+1 －L N ) Not equal to 0, andwhen determining that the current power equipment has sudden trip short circuit, wherein L N Is the sum of the sound intensities of the voiceprint signals of the current power equipment, L N+1 And L N+2 Respectively summing the sound intensities of the sound signals of the adjacent devices of the current power device;
and when the deviation degree of a certain phase is larger than a second threshold value, determining that the phase has overload faults.
4. The method for identifying a fault voiceprint of a power device by combining physical characteristics and data diagnosis according to claim 1, wherein the abnormal acoustic voiceprint signal comprises a harmonic condition and a dc bias condition, the performing the audio variation identification on the frequency domain discrete digital signal sequence, and screening out the abnormal acoustic voiceprint signal and the normal acoustic voiceprint signal comprises:
And determining that the voiceprint signal of the current power equipment is an acoustic normal voiceprint signal or a harmonic working condition or a direct current magnetic bias working condition according to the amplitude and the duty ratio of frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence.
5. The method for identifying a fault voiceprint of a power device by combining physical characteristics and data diagnosis according to claim 4, wherein determining that the voiceprint signal of the current power device is an acoustic normal voiceprint signal or a harmonic operating condition or a dc bias operating condition according to the amplitude and the duty ratio of frequency components of the voiceprint signal at different frequencies of the frequency domain discrete digital signal sequence comprises:
comparing third harmonic Y in adjacent two monitoring periods 150 Judging whether the current power equipment is in a harmonic working condition or not;
when (when)When the current power equipment is determined to be in the DC magnetic bias working condition, G M For the maximum frequency component duty ratio in the mth monitoring period, G M+1 For the maximum frequency component duty cycle in the m+1th monitoring period, D is the ratio threshold.
6. The method for identifying a power equipment fault voiceprint by combining physical characteristics and data diagnosis according to claim 1, wherein the diagnosing the primary normal voiceprint signal by using a gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal, comprises:
Learning m periods of the primary normal voiceprint signals to form a regression prediction model;
predicting the first-level normal voiceprint signal which is not learned by using the regression prediction model, and predicting the result S Y Comparing the residual K with an actual voiceprint signal Ss to form a residual K;
continuously correcting model parameters by using the residual error K, and determining model parameters of a Gaussian mixture degradation prediction algorithm;
and diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm after determining model parameters, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal.
7. The method for identifying a fault voiceprint of an electrical device by combining physical characteristics and data diagnosis according to claim 1, wherein performing fault diagnosis on the secondary abnormal voiceprint signal to determine a fault type comprises:
processing the secondary abnormal voiceprint signal by using a pre-trained neural network algorithm model to obtain a feature vector corresponding to the secondary abnormal voiceprint signal;
and comparing the feature vector corresponding to the second-level abnormal voiceprint signal with the feature vector of various faults to determine the fault type.
8. A power equipment fault voiceprint recognition system that fuses physical characteristics and data diagnostics, the system comprising:
the signal acquisition module is used for acquiring voiceprint signals of the power equipment and generating a time domain discrete digital signal sequence according to a time sequence;
the first-stage identification module is used for calculating the sound intensity variation and the sound frequency variation of the time domain discrete digital signal sequence and screening out first-stage abnormal voiceprint signals and first-stage normal voiceprint signals;
the secondary identification module is used for diagnosing the primary normal voiceprint signal by using a Gaussian mixture degradation prediction algorithm, and screening a secondary abnormal voiceprint signal and a secondary normal voiceprint signal from the primary normal voiceprint signal;
the third-level identification module is used for carrying out fault diagnosis on the second-level abnormal voiceprint signals and determining fault types;
the early warning module is used for triggering early warning based on the first-level abnormal voiceprint signal and the fault type;
the primary identification module specifically comprises:
the sound intensity recognition unit is used for recognizing the sound intensity variation degree of the time domain discrete digital signal sequence and screening out sound intensity abnormal sound track signals and sound intensity normal sound track signals;
The time-frequency domain conversion unit is used for carrying out discrete Fourier transform on the time domain discrete digital signal sequence to obtain a frequency domain discrete digital signal sequence;
the audio frequency identification unit is used for carrying out audio frequency variation identification on the frequency domain discrete digital signal sequence and screening out an abnormal audio voiceprint signal and an normal audio voiceprint signal;
a first-stage voiceprint signal determining unit configured to use the sound intensity abnormal voiceprint signal and the audio frequency abnormal voiceprint signal as the first-stage abnormal voiceprint signal, and use the sound intensity normal voiceprint signal and the audio frequency normal voiceprint signal as the first-stage normal voiceprint signal; the sound intensity recognition unit is specifically configured to:
calculating the sum of sound intensities of the voiceprint signals according to the time domain discrete digital signal sequence;
according to the sum of the sound intensities of the sound signals of the current power equipment and the sound intensities of the sound signals of the adjacent power equipment, determining that the sound signals of the current power equipment are sound intensity normal sound signals or short-time overload or long-time overload or sudden tripping short circuit;
in a power grid structure with a symmetrical A, B, C three-phase structure, determining that the voice print signal of the current power equipment is a voice print signal with normal voice print or an overload fault according to the average value of voice print signal voice print sums of the three-phase power equipment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311412959.1A CN117153193B (en) | 2023-10-30 | 2023-10-30 | Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311412959.1A CN117153193B (en) | 2023-10-30 | 2023-10-30 | Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117153193A CN117153193A (en) | 2023-12-01 |
| CN117153193B true CN117153193B (en) | 2024-02-13 |
Family
ID=88884734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311412959.1A Active CN117153193B (en) | 2023-10-30 | 2023-10-30 | Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117153193B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003050158A (en) * | 2001-08-06 | 2003-02-21 | Mitsui Eng & Shipbuild Co Ltd | Method and apparatus for diagnosis of abnormality in periodic moving body |
| JP2010134367A (en) * | 2008-12-08 | 2010-06-17 | Mitsubishi Electric Corp | Electric equipment |
| GB201210969D0 (en) * | 2011-06-27 | 2012-08-01 | Gen Electric | Electrical substation fault monitoring and diagnostics |
| WO2014114116A1 (en) * | 2013-01-28 | 2014-07-31 | Tencent Technology (Shenzhen) Company Limited | Method and system for voiceprint recognition |
| WO2016027996A1 (en) * | 2014-08-21 | 2016-02-25 | 호원대학교산학협력단 | Power equipment sound diagnosis system |
| CN114217149A (en) * | 2021-11-14 | 2022-03-22 | 国网辽宁省电力有限公司葫芦岛供电公司 | Transformer acoustic fingerprint uninterrupted power detection and state early warning method |
| CN115406522A (en) * | 2022-04-21 | 2022-11-29 | 福建华电福瑞能源发展有限公司池潭水力发电厂 | Power plant equipment running state research and application based on voiceprint recognition |
| CN116705039A (en) * | 2023-05-18 | 2023-09-05 | 江苏濠汉信息技术有限公司 | AI-based power equipment voiceprint monitoring system and method |
| CN116884432A (en) * | 2023-06-15 | 2023-10-13 | 国网冀北电力有限公司廊坊供电公司 | VMD-JS divergence-based power transformer fault voiceprint diagnosis method |
-
2023
- 2023-10-30 CN CN202311412959.1A patent/CN117153193B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003050158A (en) * | 2001-08-06 | 2003-02-21 | Mitsui Eng & Shipbuild Co Ltd | Method and apparatus for diagnosis of abnormality in periodic moving body |
| JP2010134367A (en) * | 2008-12-08 | 2010-06-17 | Mitsubishi Electric Corp | Electric equipment |
| GB201210969D0 (en) * | 2011-06-27 | 2012-08-01 | Gen Electric | Electrical substation fault monitoring and diagnostics |
| WO2014114116A1 (en) * | 2013-01-28 | 2014-07-31 | Tencent Technology (Shenzhen) Company Limited | Method and system for voiceprint recognition |
| WO2016027996A1 (en) * | 2014-08-21 | 2016-02-25 | 호원대학교산학협력단 | Power equipment sound diagnosis system |
| CN114217149A (en) * | 2021-11-14 | 2022-03-22 | 国网辽宁省电力有限公司葫芦岛供电公司 | Transformer acoustic fingerprint uninterrupted power detection and state early warning method |
| CN115406522A (en) * | 2022-04-21 | 2022-11-29 | 福建华电福瑞能源发展有限公司池潭水力发电厂 | Power plant equipment running state research and application based on voiceprint recognition |
| CN116705039A (en) * | 2023-05-18 | 2023-09-05 | 江苏濠汉信息技术有限公司 | AI-based power equipment voiceprint monitoring system and method |
| CN116884432A (en) * | 2023-06-15 | 2023-10-13 | 国网冀北电力有限公司廊坊供电公司 | VMD-JS divergence-based power transformer fault voiceprint diagnosis method |
Non-Patent Citations (2)
| Title |
|---|
| 基于声纹识别的变压器工况检测方法及验证系统;张可 等;计算技术与自动化;第1-5页 * |
| 李天匀等.船体结构振动与声学基础.华中科技大学出版社,2020,第183页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117153193A (en) | 2023-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734192B (en) | Voting mechanism-based support vector machine mechanical fault diagnosis method | |
| CN112201260B (en) | Transformer running state online detection method based on voiceprint recognition | |
| Yang et al. | Condition evaluation for opening damper of spring operated high-voltage circuit breaker using vibration time-frequency image | |
| CN112857767B (en) | Hydro-turbo generator set rotor fault acoustic discrimination method based on convolutional neural network | |
| Yang et al. | Fault identification for circuit breakers based on vibration measurements | |
| KR20200014129A (en) | Diagnosis method of electric transformer using Deep Learning | |
| CN111653291B (en) | Intelligent health monitoring method for power equipment based on voiceprint | |
| CN110231529A (en) | A kind of control cabinet intelligent Fault Diagnose Systems and method for diagnosing faults | |
| CN109190598A (en) | A kind of rotating machinery monitoring data noise detection method based on SES-LOF | |
| CN112557966A (en) | Transformer winding looseness identification method based on local mean decomposition and support vector machine | |
| CN107798283A (en) | A kind of neural network failure multi classifier based on the acyclic figure of decision-directed | |
| CN117153193B (en) | Power equipment fault voiceprint recognition method integrating physical characteristics and data diagnosis | |
| CN116796261B (en) | Closed switch equipment mechanical characteristic prediction method based on artificial intelligence | |
| CN116884432A (en) | VMD-JS divergence-based power transformer fault voiceprint diagnosis method | |
| Choudhary et al. | Fault diagnosis of induction motor under varying operating condition | |
| Dang et al. | Cochlear filter cepstral coefficients of acoustic signals for mechanical faults identification of power transformer | |
| CN116699400A (en) | Generator rotor short-circuit fault monitoring system, method and readable storage medium | |
| CN115828144A (en) | Signal sparse representation and fusion detection method, storage medium and electronic device | |
| CN112729531B (en) | Fault studying and judging method and system for distribution transformer equipment | |
| CN207181560U (en) | Transformer acoustic wave sensing system | |
| Guo et al. | Fault monitoring and diagnosis of induction machines based on harmonic wavelet transform and wavelet neural network | |
| CN116298881B (en) | Electrical signal motor health early warning method based on channel attention multi-module LMMD | |
| Das et al. | Comparison of different classifier performances for condition monitoring of induction motor using DWT | |
| CN117198328A (en) | Abnormal state identification and alarm method for dynamic audio frame of ultra-high voltage transformer | |
| Doshi et al. | Tunable q-factor wavelet transform for classifying mechanical deformations in power transformer |
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 |