CN116705055A - Substation noise monitoring method, system, equipment and storage medium - Google Patents

Substation noise monitoring method, system, equipment and storage medium Download PDF

Info

Publication number
CN116705055A
CN116705055A CN202310953789.1A CN202310953789A CN116705055A CN 116705055 A CN116705055 A CN 116705055A CN 202310953789 A CN202310953789 A CN 202310953789A CN 116705055 A CN116705055 A CN 116705055A
Authority
CN
China
Prior art keywords
noise
data
sample set
neural network
substation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202310953789.1A
Other languages
Chinese (zh)
Other versions
CN116705055B (en
Inventor
林宇锋
林志煌
李迎
陈雯
陈培铭
万紫阳
吴学超
廖志华
杨帆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
State Grid Fujian Electric Power Co Ltd
Xiamen Power Supply Co of State Grid Fujian Electric Power Co Ltd
Original Assignee
State Grid Fujian Electric Power Co Ltd
Xiamen Power Supply Co of State Grid Fujian Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by State Grid Fujian Electric Power Co Ltd, Xiamen Power Supply Co of State Grid Fujian Electric Power Co Ltd filed Critical State Grid Fujian Electric Power Co Ltd
Priority to CN202310953789.1A priority Critical patent/CN116705055B/en
Publication of CN116705055A publication Critical patent/CN116705055A/en
Application granted granted Critical
Publication of CN116705055B publication Critical patent/CN116705055B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
    • G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
    • G01D21/00—Measuring or testing not otherwise provided for
    • G01D21/02—Measuring two or more variables by means not covered by a single other subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0475—Generative networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/094—Adversarial learning
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
    • G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Remote Monitoring And Control Of Power-Distribution Networks (AREA)

Abstract

The application relates to a transformer substation noise monitoring method, a system, equipment and a storage medium, wherein the method comprises the following steps: acquiring historical environmental data and electrical data of each transformer substation, and constructing a first sample set, wherein a sample of the first sample set comprises environmental data, electrical data and an acceleration factor label in a historical period; training a first neural network based on the first sample set to obtain a trained acceleration factor prediction model; collecting audio data of a target transformer substation in the current period, and constructing a second sample set, wherein samples of the second sample set comprise the audio data and corresponding noise type labels; training a second neural network based on the second sample set, and adaptively performing sample oversampling in the process to obtain a noise equipment identification model; predicting the current acceleration factor of the target transformer substation through the acceleration factor prediction model, calculating an aging state, and retraining the second neural network when the aging state reaches a preset value; and monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training.

Description

Substation noise monitoring method, system, equipment and storage medium
Technical Field
The application relates to a transformer substation noise monitoring method, a system, equipment and a storage medium, and belongs to the technical field of noise monitoring.
Background
Substation noise refers to noise generated by certain carriers during operation of the substation, and includes transformer loop noise, unit noise, damper noise and the like. The noise has serious influence on the surrounding environment, such as hurting the eardrum of a person, affecting the normal sleep of a resident, affecting the survival of a species, and the like, so that the noise monitoring of the transformer substation is particularly important, and the equipment emitting the noise can be rapidly positioned, so that the response is timely made.
The prior art, for example, the application patent with the patent number of CN111412980A, discloses a transformer substation noise wireless monitoring system, and specifically discloses the following technical scheme: receiving noise data collected by each noise sensor; respectively extracting feature vectors of noise data acquired by each noise sensor; and inputting all the feature vectors into the trained first model, and outputting fault types. The first model is a deep neural network, a convolutional neural network or a recurrent neural network.
The prior art solves the technical problem that the fault type of noise cannot be identified, but the prior art has the problems that the frequency of noise emitted by equipment in a transformer substation is not frequent, the frequency of noise emitted by equipment of different types is greatly different, the sample of noise emitted by different equipment and the sample of noise not emitted by equipment are extremely unbalanced, so that the situation that model training is fitted is caused, and the identification precision is low; in addition, the transformer substation equipment can be aged, noise generated by each equipment is inconsistent under different aging states, and the equipment with noise cannot be accurately identified by the model at the later stage of arrangement.
Disclosure of Invention
In order to solve the problems in the prior art, the application provides a transformer substation noise monitoring method, a system, equipment and a storage medium.
The technical scheme of the application is as follows:
in one aspect, the application provides a substation noise monitoring method, which comprises the following steps:
acquiring historical environment data and electrical data of a plurality of sample substations, and constructing a first sample set, wherein each sample in the first sample set comprises environment data, electrical data and an acceleration factor label in a historical period, and the acceleration factor label is used for indicating equipment aging rate in the substation;
training a first neural network based on the first sample set to obtain a trained acceleration factor prediction model;
determining the type of noise-generating equipment contained in the target substation; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise;
training a second neural network based on the second sample set, and adaptively performing sample oversampling in the second neural network training process to obtain a noise equipment identification model;
acquiring historical environment data and electrical data of a target transformer substation from a time node to a current period, inputting the historical environment data and the electrical data into an acceleration factor prediction model, predicting acceleration factors of the target transformer substation at different historical time nodes, calculating the aging state of equipment in the target transformer substation based on the acceleration factors of the different historical time nodes, reconstructing a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and training a second neural network based on the new second sample set;
and monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training.
As a preferred embodiment, the preprocessing process of the audio data specifically includes:
performing STFT short-time Fourier transform on the audio data, performing frequency domain conversion, and calculating the frequency spectrum amplitude, wherein the method specifically comprises the following steps:
；
wherein ,is the frequency spectrum amplitude, f is the frequency axis, T is the time axis, T is the total length of the time window,/->For time window sequence number, ">Time window +.>Audio data of->The table is a window function, e is a natural constant, < ->Is the sampling frequency of the audio data;
and obtaining a power spectrum based on the frequency domain signal of the audio data, converting the power spectrum into a Mel spectrogram by a method of using a Mel filter and the calculated spectrum amplitude according to the Mel cepstrum coefficient of the MFCC, and taking the Mel spectrogram as the preprocessed audio data.
As a preferred embodiment, the method for adaptively performing oversampling on the sample in the second neural network training process specifically includes:
acquiring the classification performance of the current second neural network:
；
wherein ,indicating classification performance, N is the number of noise types, i is the ith noise type, ++>Representing the classification accuracy of the current second neural network to the ith noise type;
when classifying performanceWhen the performance threshold value is smaller than the preset performance threshold value, judging whether any current second neural network is opposite to the ith noise typeClassification accuracy of->If the accuracy is smaller than the accuracy threshold, oversampling is carried out on the sample corresponding to the noise type, otherwise, oversampling is carried out on the sample corresponding to the noise type with the minimum classification accuracy of the current second neural network;
the accuracy threshold is determined by the classification performance of the current second neural network and the maximum value of classification accuracy in all noise types, and specifically comprises the following steps:
；
wherein ,represents an accuracy threshold, ++>For a preset dimensionless scale factor, max (·) represents a maximum function.
In a preferred embodiment, in the step of adaptively performing oversampling on the sample in the second neural network training process, the method for performing oversampling is as follows:
sample oversampling is performed by Smote sampling or by using the GAN challenge-generation network as a generator.
As a preferred embodiment, in the step of calculating the aging state of the equipment in the target transformer substation based on the acceleration factors of the different historical time nodes, when the aging state of the equipment in the target transformer substation reaches a preset value, reconstructing a second sample set and training a second neural network based on the new second sample set, the method for calculating the aging state of the equipment in the target transformer substation specifically includes:
the aging percentage of the kth device was calculated:
；
wherein ,for the current aging percentage of the kth device,/->Is the initial aging percentage of the kth device,/->For the preset life cycle of the kth device, < >>Indicating the working time from the operation of the kth device to the j-time node, +.>Acceleration factor for j time nodes, j belonging to the historical time node series +.>, wherein />Time node for the device to be put into operation, +.>Is the current time node;
and respectively setting the preset value of the aging state of each device in the target substation as the percentage value of a plurality of gears which continuously ascend, and when the current aging percentage of any device in the target substation reaches any gear, starting to reconstruct the second sample set and training the second neural network.
As a preferred embodiment, the environmental data includes temperature data, humidity data, air salt content data, corrosive gas content data, rainfall data, wind speed data, and ultraviolet radiation data;
the electrical data includes average load data, line loss rate data, and three-phase imbalance rate data.
As a preferred embodiment, the first neural network employs an LSTM long short-term memory neural network, and the second neural network employs a CNN convolutional neural network.
On the other hand, the application also provides a transformer substation noise monitoring system, which comprises:
the system comprises a first sample set construction module, a second sample set construction module and a first test module, wherein the first sample set construction module is used for acquiring historical environmental data and electrical data of each transformer substation and constructing a first sample set, each sample in the first sample set comprises environmental data, electrical data and an acceleration factor label in a historical period, and the acceleration factor label is used for indicating equipment aging rate in the transformer substation;
the first model training module trains a first neural network based on the first sample set to obtain a trained acceleration factor prediction model;
the second sample set construction module is used for determining the type of equipment which generates noise and is contained in the target transformer substation; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise;
the second model training module is used for training a second neural network based on a second sample set, and in the training process of the second neural network, the sample is self-adaptively subjected to oversampling to obtain a noise equipment identification model;
the network updating module is used for acquiring environmental data and electrical data of the target transformer substation in the current period, inputting the environmental data and the electrical data into the acceleration factor prediction model, predicting the current acceleration factor of the target transformer substation, calculating the aging state of equipment in the target transformer substation based on the current acceleration factor, reconstructing a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and training a second neural network based on the new second sample set;
and the detection module is used for monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training.
In still another aspect, the present application further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the substation noise monitoring method according to any of the embodiments of the present application when the program is executed.
In yet another aspect, the present application further provides a computer readable storage medium having stored thereon a computer program which when executed by a processor implements a substation noise monitoring method according to any of the embodiments of the present application.
The application has the following beneficial effects:
according to the transformer substation noise monitoring method, system, equipment and storage medium, the aging rate of transformer substation equipment is predicted through the environmental data and the electrical data of the transformer substation, the noise type recognition model is trained through the audio data of the transformer substation, the samples are adaptively and oversampled in the training process, the balance degree of different types of samples is improved to improve the model recognition precision, the aging state is calculated according to the aging rate of the transformer substation equipment, and whether the noise equipment recognition model is required to be retrained is determined according to the aging state, so that the model can change along with the aging state of the equipment, and the recognition precision is ensured.
Drawings
Fig. 1 is a schematic flow chart of a method according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
It should be understood that the step numbers used herein are for convenience of description only and are not limiting as to the order in which the steps are performed.
It is to be understood that the terminology used in the description of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
The terms "comprises" and "comprising" indicate the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The term "and/or" refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Embodiment one:
referring to fig. 1, the present embodiment provides a substation noise monitoring method, which specifically includes the following steps:
s100, acquiring historical environment data and electrical data of a plurality of sample substations, and constructing a first sample set.
In the present embodiment, the environmental data includes temperature data, humidity data, air salt content data, corrosive gas content data, rainfall data, wind speed data, and ultraviolet radiation data; the electrical data includes average load data, line loss rate data, and three-phase imbalance rate data, where the environmental data directly or indirectly affects the aging progress of the electrical device, and for the electrical data, such as average load data affects the workload of the electrical device, the workload naturally ages fast, the workload naturally ages slow, and the average load data refers to an average load of a period of time, such as daily, weekly, and monthly.
Each sample in the first sample set contains environmental data, electrical data, and acceleration factor tags over a historical period of time, in a specific form such as T1: { A1, A2, A3, A4, A5, A6, A7, B1, B2, B3, Y1}, wherein A1, A2, A3, A4, A5, A6, A7 are respectively temperature data, humidity data, air salt content data, corrosive gas content data, rainfall data, wind speed data and ultraviolet radiation data corresponding to a history period T1, B2, B3 are respectively average load data, line loss rate data and three-phase imbalance rate data corresponding to the history period T1, Y1 is an acceleration factor label corresponding to the history period T1, the acceleration factor label is calculated and marked by an expert according to experience, and the acceleration factor label is used for indicating the aging rate of equipment in a transformer substation.
S200, based on the first sample set constructed in the step S100, taking the environmental data and the electrical data of each sample corresponding to the historical period as input, taking the acceleration factor label of the sample as output to perform iterative training on the first neural network, setting iteration termination conditions, for example, the error between a predicted value and a real label value is less than 1%, and storing the current network parameters after the iteration is completed to obtain a trained acceleration factor prediction model.
S300, determining the type of equipment contained in the target substation and generating noise, such as a transformer, a high-voltage reactor, a cable and the like; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise;
s400, training a second neural network based on the second sample set constructed in the step S300, taking the preprocessed audio data as the input of the network and the noise type label as the output of the network in the training process of the second neural network, and adaptively performing sample oversampling at the same time, wherein the purpose of the oversampling is to ensure that samples of all noise types are relatively balanced, so that the second neural network cannot be subjected to line-out and fitting conditions, and also set iteration termination conditions, and storing current network parameters after iteration is completed to obtain a noise equipment identification model.
S500, acquiring historical environment data and electrical data of a target transformer substation from a time node of operation to a current period, wherein the data can be easily obtained from a weather prediction system and a power grid management system, are input into an acceleration factor prediction model, predict acceleration factors of the target transformer substation at different time nodes, calculate the aging state of equipment in the target transformer substation based on the acceleration factors of the different time nodes, reconstruct a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and train a second neural network based on the new second sample set;
and S600, monitoring the noise of the transformer substation based on the noise equipment identification model obtained through current training.
Based on the embodiment, by taking historical environmental data and electrical data of different substations as samples, an acceleration factor prediction model capable of predicting the aging rate of equipment in the transformer substation according to the environmental data and the electrical data of different time periods is trained, then noise equipment recognition models capable of recognizing different noise equipment are trained by using noise audio data, and the recognition precision of the noise equipment recognition models is improved by using an oversampling means; meanwhile, the ageing state of the equipment in the transformer substation is calculated through the predicted ageing rate of the equipment in the transformer substation, and when the ageing state of the equipment reaches a certain degree, the noise equipment identification model is retrained so as to ensure that the noise equipment identification model can be updated and optimized along with the ageing degree of the equipment, so that the noise equipment identification model cannot lose identification precision due to the change of noise audio caused by equipment ageing, and the effectiveness and the accuracy of transformer substation noise monitoring are greatly improved.
As a preferred implementation manner of this embodiment, the preprocessing process of the audio data specifically includes:
performing STFT short-time Fourier transform on the audio data, performing frequency domain conversion, and calculating the frequency spectrum amplitude, wherein the method specifically comprises the following steps:
；
wherein ,is the frequency spectrum amplitude, f is the frequency axis, T is the time axis, T is the total length of the time window,/->For time window sequence number, ">Time window +.>Audio data of (a)，/>The table is a window function, e is a natural constant, < ->Is the sampling frequency of the audio data;
obtaining a power spectrum based on a frequency domain signal of the audio data, converting the power spectrum into a Mel spectrogram by a method of using a Mel filter and the calculated spectrum amplitude according to the MFCC Mel cepstrum coefficient, and taking the Mel spectrogram as the preprocessed audio data;
in order to solve the problem that the original audio data is one-dimensional information, only the time domain information of sound can be acquired, and the frequency domain information cannot be acquired, the embodiment performs short-time fourier transform (STFT) on the original audio data, that is, performs fourier transform on the short-time audio signal: framing and windowing an audio signal with a certain length, performing Fourier transform on each frame, and stacking the result of each frame along another dimension to obtain a two-dimensional spectrogram; meanwhile, as the feature dimension of the input two-dimensional spectrogram is too large to perform feature extraction, the original spectrogram is converted into a Mel spectrogram, and because the Mel spectrogram adopts Mel scales, the specific features can be reserved for human ears, the input feature dimension can be ensured to be more consistent with the features of noise which can be detected by the human ears, and the accuracy of a noise equipment identification model is improved.
As a preferred implementation manner of this embodiment, the method for adaptively performing oversampling on the sample in the second neural network training process specifically includes:
acquiring the classification performance of the current second neural network:
；
wherein ,representing classification performance; n is noiseThe number of types, i.e. the number of noise-generating device types contained in the determined target substation; i denotes the i-th noise type, i.e. the device type; />Representing the classification accuracy of the current second neural network to the ith noise type;
when classifying performanceLess than a preset performance threshold->When, for example, the performance threshold is preset +.>95% when the classification performance of the second neural network is present +.>If the classification accuracy of the current second neural network to the ith noise type is less than 95%, judging whether the classification accuracy of any current second neural network to the ith noise type is +.>Less than the accuracy threshold->For example, there are 5 noise types currently, and the classification accuracy of the second neural network to the second neural network is respectively F1 (98%), F2 (97%), F3 (93%), F4 (88%), F5 (96%), so that the classification performance ∈can be calculated>94.4% less than the predetermined performance threshold, starting to determine if the classification accuracy of any noise type is less than the accuracy threshold, and if so, oversampling the samples of the corresponding noise type, in which case if the predetermined accuracy threshold is 94%, both F3 and F4 are less than the accuracy threshold, and both samples of the noise type corresponding to F3 and F4 need to be oversampled, if the predetermined accuracy threshold is 90%Only the noise type samples corresponding to F4 need be oversampled.
If the classification accuracy of all the noise types is greater than or equal to the accuracy threshold, only the sample corresponding to the noise type with the minimum classification accuracy of the current second neural network is needed to be oversampled;
in this embodiment, the accuracy threshold is adaptively adjusted, specifically determined by the classification performance of the current second neural network and the maximum value of classification accuracy among all noise types, specifically:
；
wherein ,represents an accuracy threshold, ++>For a preset dimensionless scale factor, max (·) represents a maximum function.
As a preferred implementation manner of this embodiment, in the step of adaptively performing oversampling on a sample in the second neural network training process, the method for performing oversampling is:
sample oversampling is carried out by a Smote sampling method, or a GAN countermeasure generation network is used as a generator to carry out sample oversampling;
specifically, the present embodiment takes the GAN countermeasure generation network as a generator to sample oversampling, for the GAN countermeasure generation network:
；
wherein ,for a valued function, ++>Representation when sample->Belonging to->The time value is 1 @, @>Representation when sample->Belonging to->The value is 0; />= { xr1, xr2, … …, xrn } represents a real data set, wherein xrn represents a real sample of the nth input, i.e., the nth preprocessed audio data; />= { xf1, xf2, … …, xfm } represents the generated dataset, where xfm represents the mth generated sample;
the loss function of the GAN challenge-generating network is set as follows:
；
wherein ,data representing the generation of a potential vector z conforming to a gaussian distribution, gamma being a preset gradient penalty term,/->For gradient operator->Representing a set of potential vectors z;
training the GAN countermeasure generation network based on the loss function and the real sample, obtaining a trained network as a generator, and performing sample oversampling through the generator.
Based on the above embodiment, in a general GAN network, it is desirable that the Loss is as large as possible to determine the difference between the true and false samples, which results in that the gradient calculated by Loss in the discriminator will change along the direction of increasing Loss, but after interception, each network parameter is independently limited in the value range. This result makes all parameters extreme, either taking the maximum value, e.g. 0.01, or the lowest value (e.g. -0.01), the arbiter fails to fully and model itself, and the gradient passed through to the generator will then be degraded. The penalty function of this embodiment sets a gradient penalty term, and when the L2 regularization of the gradient has a value other than 1, the rest of the Loss is optimized by giving as much penalty as possible, thus solving the problems of gradient penalty and mode collapse.
As a preferred implementation manner of this embodiment, in the step of calculating the aging state of the device in the target substation based on the acceleration factors of the different historical time nodes, when the aging state of the device in the target substation reaches a preset value, reconstructing the second sample set and training the second neural network based on the new second sample set, the method for calculating the aging state of the device in the target substation specifically includes:
the aging percentage of the kth device was calculated:
；
wherein ,for the current aging percentage of the kth device,/->Is the initial aging percentage of the kth device,/->For the preset life cycle of the kth device, < >>Indicating the working time from the operation of the kth device to the j-time node, +.>Acceleration factor for j time nodes, j belonging to the historical time node series +.>, wherein />Time node for the device to be put into operation, +.>Is the current time node;
the preset values of the aging states of the devices in the target transformer substation are respectively set to be percentage values of a plurality of continuously ascending gears, for example, the gear is set to be (10%, 20%,40%, 70%) for the device A, and the gear is set to be (10%, 20%,30%,40%, 60%) for the device B
When the current aging percentage of any device in the target substation reaches any gear, as exemplified above, when the aging state of the device A reaches 10%, the second sample set is reconstructed and the second neural network is trained, and when the aging state of the device B reaches 20%, the second sample set is reconstructed and the second neural network is trained.
Based on the above embodiment, the embodiment uses the aging state of the device as the basis for reconstructing and training the second neural network, and retrains the model only when the device does change due to aging, so that the running time of model training is most valuable, and the computational power resource requirement is reduced.
As a preferred implementation manner of this embodiment, the first neural network adopts an LSTM long and short-term memory neural network, where the samples in the first sample set are sequence data, and the LSTM long and short-term memory neural network is a commonly used neural network for processing the sequence data; the second neural network adopts a CNN convolutional neural network, and samples in the sample set are Mel spectrograms, so that image features can be captured through the CNN convolutional neural network well.
Embodiment two:
the embodiment provides a transformer substation noise monitoring system, including:
the system comprises a first sample set construction module, a second sample set construction module and a first test module, wherein the first sample set construction module is used for acquiring historical environmental data and electrical data of each transformer substation and constructing a first sample set, each sample in the first sample set comprises environmental data, electrical data and an acceleration factor label in a historical period, and the acceleration factor label is used for indicating equipment aging rate in the transformer substation; the module is used for implementing the function of step S100 in the first embodiment, and will not be described here again;
the first model training module trains a first neural network based on the first sample set to obtain a trained acceleration factor prediction model; the module is used for implementing the function of step S200 in the first embodiment, and will not be described in detail herein;
the second sample set construction module is used for determining the type of equipment which generates noise and is contained in the target transformer substation; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise; the module is used for implementing the function of step S300 in the first embodiment, and will not be described in detail herein;
the second model training module is used for training a second neural network based on a second sample set, and in the training process of the second neural network, the sample is self-adaptively subjected to oversampling to obtain a noise equipment identification model; the module is used for realizing the function of step S400 in the first embodiment, and will not be described in detail herein;
the network updating module is used for acquiring environmental data and electrical data of the target transformer substation in the current period, inputting the environmental data and the electrical data into the acceleration factor prediction model, predicting the current acceleration factor of the target transformer substation, calculating the aging state of equipment in the target transformer substation based on the current acceleration factor, reconstructing a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and training a second neural network based on the new second sample set; the module is used for realizing the function of step S500 in the first embodiment, and will not be described in detail herein;
the detection module is used for monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training; the module is used to implement the function of step S600 in the first embodiment, which is not described herein.
Embodiment III:
the embodiment provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the transformer substation noise monitoring method according to any embodiment of the application when executing the program.
Embodiment four:
the present embodiment proposes a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a substation noise monitoring method according to any of the embodiments of the present application.
In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "and/or", describes an association relation of association objects, and indicates that there may be three kinds of relations, for example, a and/or B, and may indicate that a alone exists, a and B together, and B alone exists. Wherein A, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of the following" and the like means any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c may represent: a, b, c, a and b, a and c, b and c or a and b and c, wherein a, b and c can be single or multiple.
Those of ordinary skill in the art will appreciate that the various elements and algorithm steps described in the embodiments disclosed herein can be implemented as a combination of electronic hardware, computer software, and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In several embodiments provided by the present application, any of the functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (hereinafter referred to as ROM), a random access Memory (Random Access Memory) and various media capable of storing program codes such as a magnetic disk or an optical disk.
The foregoing description is only illustrative of the present application and is not intended to limit the scope of the application, and all equivalent structures or equivalent processes or direct or indirect application in other related technical fields are included in the scope of the present application.

Claims (10)

1. The transformer substation noise monitoring method is characterized by comprising the following steps of:
acquiring historical environment data and electrical data of a plurality of sample substations, and constructing a first sample set, wherein each sample in the first sample set comprises environment data, electrical data and an acceleration factor label in a historical period, and the acceleration factor label is used for indicating equipment aging rate in the substation;
training a first neural network based on the first sample set to obtain a trained acceleration factor prediction model;
determining the type of noise-generating equipment contained in the target substation; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise;
training a second neural network based on the second sample set, and adaptively performing sample oversampling in the second neural network training process to obtain a noise equipment identification model;
acquiring historical environment data and electrical data of a target transformer substation from a time node to a current period, inputting the historical environment data and the electrical data into an acceleration factor prediction model, predicting acceleration factors of the target transformer substation at different historical time nodes, calculating the aging state of equipment in the target transformer substation based on the acceleration factors of the different historical time nodes, reconstructing a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and training a second neural network based on the new second sample set;
and monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training.
2. The substation noise monitoring method according to claim 1, wherein the preprocessing process of the audio data specifically comprises:
performing STFT short-time Fourier transform on the audio data, performing frequency domain conversion, and calculating the frequency spectrum amplitude, wherein the method specifically comprises the following steps:
；
wherein ,is the frequency spectrum amplitude, f is the frequency axis, T is the time axis, and T is the total length of the time windowDegree (f)>For time window sequence number, ">Time window +.>Audio data of->The table is a window function, e is a natural constant, < ->Is the sampling frequency of the audio data;
and obtaining a power spectrum based on the frequency domain signal of the audio data, converting the power spectrum into a Mel spectrogram by a method of using a Mel filter and the calculated spectrum amplitude according to the Mel cepstrum coefficient of the MFCC, and taking the Mel spectrogram as the preprocessed audio data.
3. The substation noise monitoring method according to claim 1, wherein the method for adaptively performing oversampling of samples during the second neural network training process specifically comprises:
acquiring the classification performance of the current second neural network:
；
wherein ,indicating classification performance, N is the number of noise types, i is the ith noise type, ++>Representing the current second neural network versus the ith noise typeClassification accuracy of (2);
when classifying performanceWhen the classification accuracy of any one of the current second neural networks to the ith noise type is less than the preset performance threshold value, judging whether the classification accuracy of any one of the current second neural networks to the ith noise type is +.>If the accuracy is smaller than the accuracy threshold, oversampling is carried out on the sample corresponding to the noise type, otherwise, oversampling is carried out on the sample corresponding to the noise type with the minimum classification accuracy of the current second neural network;
the accuracy threshold is determined by the classification performance of the current second neural network and the maximum value of classification accuracy in all noise types, and specifically comprises the following steps:
；
wherein ,represents an accuracy threshold, ++>For a preset dimensionless scale factor, max (·) represents a maximum function.
4. The substation noise monitoring method according to claim 1, wherein in the step of adaptively performing oversampling of the samples in the second neural network training process, the method for performing oversampling is:
sample oversampling is performed by Smote sampling or by using the GAN challenge-generation network as a generator.
5. The substation noise monitoring method according to claim 1, wherein in the step of calculating the aging state of the equipment in the target substation based on the acceleration factors of the different historical time nodes, when the aging state of the equipment in the target substation reaches a preset value, reconstructing the second sample set and training the second neural network based on the new second sample set, the method for calculating the aging state of the equipment in the target substation specifically comprises:
the aging percentage of the kth device was calculated:
；
wherein ,for the current aging percentage of the kth device,/->Is the initial aging percentage of the kth device,/->For the preset life cycle of the kth device, < >>Indicating the working time from the operation of the kth device to the j-time node, +.>Acceleration factor for j time nodes, j belonging to the historical time node series +.>, wherein />Time node for the device to be put into operation, +.>Is the current time node;
and respectively setting the preset value of the aging state of each device in the target substation as the percentage value of a plurality of gears which continuously ascend, and when the current aging percentage of any device in the target substation reaches any gear, starting to reconstruct the second sample set and training the second neural network.
6. A substation noise monitoring method according to claim 1, characterized by:
the environment data comprise temperature data, humidity data, air salt content data, corrosive gas content data, rainfall data, wind speed data and ultraviolet radiation data;
the electrical data includes average load data, line loss rate data, and three-phase imbalance rate data.
7. A substation noise monitoring method according to claim 1, characterized by:
the first neural network adopts an LSTM long short-time memory neural network, and the second neural network adopts a CNN convolutional neural network.
8. A substation noise monitoring system, comprising:
the system comprises a first sample set construction module, a second sample set construction module and a first test module, wherein the first sample set construction module is used for acquiring historical environmental data and electrical data of each transformer substation and constructing a first sample set, each sample in the first sample set comprises environmental data, electrical data and an acceleration factor label in a historical period, and the acceleration factor label is used for indicating equipment aging rate in the transformer substation;
the first model training module trains a first neural network based on the first sample set to obtain a trained acceleration factor prediction model;
the second sample set construction module is used for determining the type of equipment which generates noise and is contained in the target transformer substation; collecting audio data of a target transformer substation in a current period, and constructing a second sample set, wherein each sample in the second sample set comprises preprocessed audio data and a corresponding noise type label, and the noise type label indicates the type of equipment generating noise;
the second model training module is used for training a second neural network based on a second sample set, and in the training process of the second neural network, the sample is self-adaptively subjected to oversampling to obtain a noise equipment identification model;
the network updating module is used for acquiring environmental data and electrical data of the target transformer substation in the current period, inputting the environmental data and the electrical data into the acceleration factor prediction model, predicting the current acceleration factor of the target transformer substation, calculating the aging state of equipment in the target transformer substation based on the current acceleration factor, reconstructing a second sample set when the aging state of the equipment in the target transformer substation reaches a preset value, and training a second neural network based on the new second sample set;
and the detection module is used for monitoring the noise of the transformer substation based on the noise equipment identification model obtained by the current training.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the substation noise monitoring method of any of claims 1 to 7 when the program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the substation noise monitoring method according to any one of claims 1-7.
CN202310953789.1A 2023-08-01 2023-08-01 Substation noise monitoring method, system, equipment and storage medium Active CN116705055B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310953789.1A CN116705055B (en) 2023-08-01 2023-08-01 Substation noise monitoring method, system, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310953789.1A CN116705055B (en) 2023-08-01 2023-08-01 Substation noise monitoring method, system, equipment and storage medium

Publications (2)

Publication Number Publication Date
CN116705055A true CN116705055A (en) 2023-09-05
CN116705055B CN116705055B (en) 2023-10-17

Family

ID=87832459

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310953789.1A Active CN116705055B (en) 2023-08-01 2023-08-01 Substation noise monitoring method, system, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN116705055B (en)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107705805A (en) * 2017-10-25 2018-02-16 北京奇虎科技有限公司 Audio duplicate checking method and device
CN110136744A (en) * 2019-05-24 2019-08-16 腾讯音乐娱乐科技（深圳）有限公司 A kind of audio-frequency fingerprint generation method, equipment and storage medium
CN110910891A (en) * 2019-11-15 2020-03-24 复旦大学 Speaker segmentation labeling method and device based on long-time memory neural network
WO2021000498A1 (en) * 2019-07-04 2021-01-07 平安科技（深圳）有限公司 Composite speech recognition method, device, equipment, and computer-readable storage medium
CN112509600A (en) * 2020-12-11 2021-03-16 平安科技（深圳）有限公司 Model training method and device, voice conversion method and device and storage medium
US20220147824A1 (en) * 2020-11-12 2022-05-12 equal1.labs Inc. Accelerated Learning In Neural Networks Incorporating Quantum Unitary Noise And Quantum Stochastic Rounding Using Silicon Based Quantum Dot Arrays
WO2022140794A1 (en) * 2020-12-23 2022-06-30 Lucomm Technologies, Inc. Flux sensing system

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107705805A (en) * 2017-10-25 2018-02-16 北京奇虎科技有限公司 Audio duplicate checking method and device
CN110136744A (en) * 2019-05-24 2019-08-16 腾讯音乐娱乐科技（深圳）有限公司 A kind of audio-frequency fingerprint generation method, equipment and storage medium
WO2021000498A1 (en) * 2019-07-04 2021-01-07 平安科技（深圳）有限公司 Composite speech recognition method, device, equipment, and computer-readable storage medium
CN110910891A (en) * 2019-11-15 2020-03-24 复旦大学 Speaker segmentation labeling method and device based on long-time memory neural network
US20220147824A1 (en) * 2020-11-12 2022-05-12 equal1.labs Inc. Accelerated Learning In Neural Networks Incorporating Quantum Unitary Noise And Quantum Stochastic Rounding Using Silicon Based Quantum Dot Arrays
CN112509600A (en) * 2020-12-11 2021-03-16 平安科技（深圳）有限公司 Model training method and device, voice conversion method and device and storage medium
WO2022140794A1 (en) * 2020-12-23 2022-06-30 Lucomm Technologies, Inc. Flux sensing system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李小龙: "钻井井场噪声采集与处理技术研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》, no. 9, pages 10 - 21 *

Also Published As

Publication number Publication date
CN116705055B (en) 2023-10-17

Similar Documents

Publication Publication Date Title
CN111445905B (en) Mixed voice recognition network training method, mixed voice recognition method, device and storage medium
CN115412455B (en) Method and device for detecting abnormality of multiple performance indexes of server based on time sequence
CN111815806B (en) Method for preprocessing flight parameter data based on wild value elimination and feature extraction
JP2020126021A (en) Abnormal sound detector and abnormal sound detection method
CN108630209B (en) Marine organism identification method based on feature fusion and deep confidence network
CN110175541B (en) Method for extracting sea level change nonlinear trend
CN109884419A (en) A kind of wisdom grid power quality on-line fault diagnosis method
CN105678343A (en) Adaptive-weighted-group-sparse-representation-based diagnosis method for noise abnormity of hydroelectric generating set
CN115225516B (en) LSSVM network flow prediction method based on improved ABC-VMD
CN107689059A (en) The abnormal recognition methods of wind generating set pitch control and device
CN113111786A (en) Underwater target identification method based on small sample training image convolutional network
CN116609440A (en) Intelligent acceptance management method and system for building engineering quality based on cloud edge cooperation
CN117473414A (en) Bearing fault position identification method based on low-noise time-frequency image
Tran et al. Denoising induction motor sounds using an autoencoder
CN117688422B (en) Under-modeling state parameter identification method based on improved sparse component analysis, computer equipment and storage medium
CN117782198B (en) Highway electromechanical equipment operation monitoring method and system based on cloud edge architecture
CN110222386A (en) A kind of planetary gear degenerate state recognition methods
CN113758709A (en) Rolling bearing fault diagnosis method and system combining edge calculation and deep learning
CN113657244A (en) Fan gearbox fault diagnosis method and system based on improved EEMD and speech spectrum analysis
CN116705055B (en) Substation noise monitoring method, system, equipment and storage medium
CN113361782A (en) Photovoltaic power generation power short-term rolling prediction method based on improved MKPLS
CN116680556A (en) Method for extracting vibration signal characteristics and identifying state of water pump unit
CN115563480A (en) Gear fault identification method for screening octave geometric modal decomposition based on kurtosis ratio coefficient
CN116046396A (en) Bearing fault diagnosis method, device, equipment and medium
CN110020685A (en) A kind of preprocess method based on adaptive-filtering and limited Boltzmann machine, terminal and readable storage medium storing program for executing

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