Vibration noise evaluation method and device of power equipment, medium and equipment
By decomposing the historical load data of power equipment and extracting the sound pressure information feature of sound pressure information, combined with the evaluation model, the problem of inaccurate low-frequency noise evaluation in the existing technology is solved, the accuracy of noise evaluation is improved, and the noise reduction of power equipment is provided. Basics, optimized construction.
Patent Information
- Application Number
- CN202411047043.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art cannot accurately evaluate the low-frequency noise of substations and ignores load characteristics, resulting in a decrease in the accuracy of noise measurement.
By obtaining the historical load data of power equipment for decomposition, constructing historical daily load time series data, and collecting sound pressure information for feature extraction, combining the evaluation model to mine the relationship between load and noise from load, noise and time dimensions to obtain accurate low-frequency noise Evaluation results.
It improves the accuracy of noise evaluation, conforms to the actual operation of power equipment, provides a foundation for noise reduction, and optimizes the construction of power equipment.
Smart Images

Figure CN119004172A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of noise assessment, and in particular to a vibration noise assessment method, device, medium and equipment for electric power equipment. Background Art
[0002] With the rapid development of power systems and urban construction, the demand for electricity load in society is growing, and the symbiosis of substation construction and production and life has become a new substation construction model. Although this model can improve the power supply quality and meet the growing demand for electricity load, it also brings noise problems.
[0003] Since the transformer in the substation will vibrate during operation, it will induce serious structural vibration and secondary noise problems, and the low-frequency noise will have a greater negative impact on the normal life and health of residents. In the existing noise management methods for substations, usually only the sound of the entire substation is measured and rated, so as to make corresponding noise reduction measures. However, this method ignores the diversity of the sound sources of the substation, and thus cannot obtain accurate low-frequency noise information. In addition, since different load characteristics represent the working efficiency of the substation and the intensity of the noise, the existing sound measurement process usually only focuses on how to measure, while ignoring the load characteristics of the substation, which reduces the accuracy of the sound measurement to a certain extent. Therefore, there is an urgent need for a new method to solve the defects in the existing technology. Summary of the invention
[0004] The present invention provides a vibration noise assessment method, device, medium and equipment for electric power equipment to solve the technical problem that the low-frequency noise of a substation cannot be accurately assessed in the prior art.
[0005] In order to solve the above technical problems, an embodiment of the present invention provides a vibration noise evaluation method for electric equipment, comprising:
[0006] Obtain historical load data of power equipment and decompose it to obtain historical daily load data;
[0007] According to a preset clustering algorithm, historical load time series data is constructed in combination with the historical daily load data;
[0008] According to a preset collection method, the sound pressure information of the electric equipment is collected to obtain a standard sound pressure signal;
[0009] Extracting features of the standard sound pressure signal to obtain first noise data;
[0010] Inputting the historical load time series data and the first noise data into a preset initial evaluation model, and obtaining an optimal evaluation model according to a preset evaluation training method;
[0011] The daily load data of the power equipment to be evaluated is obtained and input into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
[0012] It can be understood that, compared with the prior art, the present invention decomposes the historical load data of the power equipment into the historical daily load data that is more closely related to the daily operation of the power equipment, and constructs the corresponding time series data based on this; by collecting the sound pressure information of the power equipment and performing feature extraction, accurate low-frequency noise data is obtained. Through the evaluation model, the relationship between load and noise is mined from the three dimensions of load, noise and time, so that the evaluation of low-frequency noise data can be obtained based on the daily load data that is most closely related to the power equipment, so that the evaluation result is consistent with the actual operation of the power equipment, the accuracy of the noise evaluation is improved, and a basis is provided for the noise reduction of the power equipment, thereby optimizing the construction of the power equipment.
[0013] As a preferred solution, the historical load data of the power equipment is decomposed to obtain the historical daily load data, specifically including:
[0014] Acquire historical load data of the power equipment, and perform data cleaning on the historical load data to obtain first historical load data;
[0015] According to the preset date feature type, combined with the Fourier decomposition method, the first historical load data is decomposed to obtain the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component;
[0016] According to a preset weighted average algorithm, the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component are weightedly counted to obtain historical daily load data.
[0017] By decomposing historical load data into historical daily load data, we can obtain load data that is more consistent with the normal working state of the power equipment, and use daily load data to better reflect the working characteristics of the power equipment.
[0018] As a preferred solution, the method of constructing the historical load time series data based on the preset clustering algorithm and combining the historical daily load data specifically includes:
[0019] The historical daily load data is converted into historical daily load matrix data, and a kernel principal component analysis algorithm is used to reduce the dimension of the high-dimensional historical daily load matrix data to obtain first historical daily load matrix data;
[0020] Transform the load matrix data of the first historical day to obtain the historical daily load curve;
[0021] According to a preset clustering algorithm, the optimal clustering result is obtained in combination with the historical daily load curve;
[0022] According to the optimal clustering result, historical load time series data is constructed.
[0023] By converting the historical daily load data into historical daily load matrix data and performing dimensionality reduction processing, and then using a clustering algorithm, the load can be classified according to date characteristics, thereby obtaining a more accurate relationship between load and time.
[0024] As a preferred solution, the sound pressure information of the electric equipment is collected according to a preset collection method to obtain a standard sound pressure signal, which specifically includes:
[0025] The sound field is constructed with the power equipment as the center point, and the power equipment is decomposed into a plurality of collection surfaces; each collection surface is decomposed into a plurality of collection points, and the sound pressure signal of each collection point and collection surface is collected, and the total sound pressure signal of the power equipment is constructed according to the sound pressure signal of each collection point and collection surface:
[0026]
[0027] p ij (r)=ρ0ω∫ D q(r0)G(r,r0),
[0028]
[0029]
[0030] Among them, p i is the sound pressure signal of the ith collection surface, A i (ω) is the acoustic transfer vector of the i-th acquisition surface, N is the number of acquisition points in the i-th acquisition surface; p ij (r) is the acoustic pressure signal of the jth acquisition point in the i-th acquisition surface; ρ0 is the average density of the medium; ω is the vibration angular frequency of the power equipment; D is the boundary of the sound field; q(r0) is the pressure generated by the vibration source of the power equipment, G(r,r0) is the Green function; r0 is the center point of the vibration source of the power equipment; r is the distance between the acquisition point j and the center point of the vibration source of the power equipment; m is the number of acquisition surfaces, v i (ω) is the vibration rate of the power equipment at the i-th collection surface; P is the sound pressure signal of the power equipment;
[0031] The sound pressure signal of the electric equipment is normalized to obtain a standard sound pressure signal of the electric equipment.
[0032] By constructing the sound field of the power equipment and decomposing it into multiple collection surfaces and collection points, a more comprehensive and accurate sound pressure signal can be collected, thereby obtaining a more accurate sound pressure signal of the power equipment. Finally, through normalization processing, large differences between signals of different levels of sound pressure signals can be avoided, thereby ensuring that each sound pressure signal collected is used for the same sound pressure level.
[0033] As a preferred solution, the normalizing the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment specifically includes:
[0034] Obtain a preset normalization ratio, and use the DBFS normalization method to normalize the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment:
[0035] DBFS max =20log 10 (max|soundsignal old |),
[0036] DBFS min = = 20log 10 (max|soundsignal old |),
[0037]
[0038] Among them, DBFS new is the maximum normalized ratio, DBFS min is the minimum normalized ratio, soundsignal old is the sound pressure signal of the power equipment before normalization, soundsignal new is the standard sound pressure signal of the power equipment obtained after normalization.
[0039] By normalizing, large differences between signals of different sound pressure levels can be avoided, thereby ensuring that each sound pressure signal collected is used for the same sound pressure level.
[0040] As a preferred solution, the feature extraction of the standard sound pressure signal to obtain the first noise data specifically includes:
[0041] Performing signal enhancement on the standard sound pressure signal according to a time shift operation method to obtain a first standard sound pressure signal;
[0042] Performing time-frequency analysis on the first standard sound pressure signal according to a short-time Fourier transform method to obtain a first time-frequency spectrum;
[0043] According to a preset feature extraction method, feature extraction is performed on the first time-frequency spectrum to obtain corresponding amplitude spectrum features and phase spectrum features;
[0044] Convolutionally encoding the amplitude spectrum features and the phase spectrum features to obtain audio features in a high-dimensional feature space;
[0045] Extracting audio features in the high-dimensional feature space according to a preset time-frequency convolution method to obtain time-frequency features;
[0046] Performing phase decoding on the time-frequency feature to obtain a corresponding phase feature;
[0047] According to a preset feature screening method, the time-frequency features and phase features are screened to obtain an audio feature set;
[0048] According to the inverse Fourier transform method, initial noise data is obtained in combination with the audio feature set;
[0049] The initial noise data is filtered according to a low-pass filtering method to obtain first noise data.
[0050] The sound pressure signal is converted into a time-frequency spectrum, and convolution coding, phase decoding, etc. are performed, and finally feature screening and low-pass filtering are performed to obtain accurate low-frequency noise data.
[0051] As a preferred solution, the historical load time series data and the first noise data are input into a preset initial evaluation model, and an optimal evaluation model is obtained according to a preset evaluation training method, which specifically includes:
[0052] Inputting the historical load time series data and the first noise data into a preset convolutional neural network model, so that the convolutional neural network model can fit the historical load time series data and the first noise data in the time dimension;
[0053] A weighted error evaluation method is set in the output layer of the convolutional neural network model to enable the evaluation result of the convolutional neural network model output for the noise data;
[0054] According to the evaluation results, the convolutional neural network model is iteratively trained, and the evaluation results are updated until the evaluation results meet the preset standards, the training of the convolutional neural network model is completed, and the optimal evaluation model is obtained.
[0055] By inputting the historical load time series data and the first noise data into the preset convolutional neural network model and adding the A-weighted error evaluation method to the convolutional neural network model, the relationship between the load data and the noise data in the time series can be fully explored and the noise level can also be evaluated.
[0056] Accordingly, an embodiment of the present invention provides a vibration and noise evaluation device for electric power equipment, comprising: a load data acquisition module, a load time series data acquisition module, a sound pressure signal acquisition module, a noise data acquisition module, an evaluation model acquisition module, and a vibration and noise evaluation module;
[0057] The load data acquisition module is used to acquire historical load data of power equipment and decompose it to obtain historical daily load data;
[0058] The load time series data acquisition module is used to construct historical load time series data based on a preset clustering algorithm and in combination with the historical daily load data;
[0059] The sound pressure signal acquisition module is used to collect the sound pressure information of the electric equipment according to a preset collection method to obtain a standard sound pressure signal;
[0060] The noise data acquisition module is used to extract features from the standard sound pressure signal to obtain first noise data;
[0061] The evaluation model acquisition module is used to input the historical load time series data and the first noise data into a preset initial evaluation model, and obtain an optimal evaluation model according to a preset evaluation training method;
[0062] The vibration and noise evaluation module is used to obtain the daily load data of the power equipment to be evaluated, and input it into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
[0063] It can be understood that, compared with the prior art, this device decomposes the historical load data of the power equipment into historical daily load data that is more closely related to the daily operation of the power equipment, and constructs the corresponding time series data based on this; by collecting the sound pressure information of the power equipment and performing feature extraction, accurate low-frequency noise data is obtained. Through the evaluation model, the relationship between load and noise is mined from the three dimensions of load, noise and time, so that the evaluation of low-frequency noise data can be obtained based on the daily load data that is most closely related to the power equipment, so that the evaluation results are consistent with the actual operation of the power equipment, the accuracy of the noise evaluation is improved, and a basis is provided for the noise reduction of the power equipment, thereby optimizing the construction of the power equipment.
[0064] Accordingly, an embodiment of the present invention provides a terminal device, including:
[0065] one or more processors;
[0066] A memory, coupled to the processor, for storing one or more programs;
[0067] When the one or more programs are executed by the one or more processors, the one or more processors implement the vibration noise evaluation method of the electric power equipment as described above.
[0068] Accordingly, an embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. The computer program is executed by a processor to implement the vibration noise evaluation method of an electric power equipment as described above. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] Figure 1 : A flowchart of a method for evaluating vibration noise of electric equipment provided by an embodiment of the present invention;
[0070] Figure 2 : A schematic diagram of the structure of a vibration noise evaluation device for electric equipment provided by an embodiment of the present invention;
[0071] Among them, 201: load data acquisition module; 202: load time series data acquisition module; 203: sound pressure signal acquisition module; 204: noise data acquisition module; 205: evaluation model acquisition module; 206: vibration noise evaluation module. DETAILED DESCRIPTION
[0072] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0073] Embodiment 1
[0074] Please refer to Figure 1 , which is a flow chart of the steps of a vibration noise evaluation method for electric equipment provided by an embodiment of the present invention, including steps S101-S106.
[0075] Step S101: Obtain historical load data of power equipment and decompose it to obtain historical daily load data.
[0076] In this embodiment, the historical load data of the power equipment is decomposed to obtain the historical daily load data, which specifically includes:
[0077] Acquire historical load data of the power equipment, and perform data cleaning on the historical load data to obtain first historical load data;
[0078] According to the preset date feature type, combined with the Fourier decomposition method, the first historical load data is decomposed to obtain the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component;
[0079] According to a preset weighted average algorithm, the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component are weightedly counted to obtain historical daily load data.
[0080] By decomposing historical load data into historical daily load data, we can obtain load data that is more consistent with the normal working state of the power equipment, and use daily load data to better reflect the working characteristics of the power equipment.
[0081] In an optional embodiment, the data cleaning of the historical load data further includes supplementing the missing data. The specific steps of supplementing the missing data include:
[0082]
[0083] i=[rand()],
[0084] n={24,14},
[0085] Among them, L t Indicates the missing value at the current time t, L t-1 represents the power load collection value at time t-1, L t+n Indicates the power load sampling value at time t+n, L t+n represents the power load sampling value at time t+i, i=[rand()] represents the value of i obtained by taking a random real number between 0 and 1 and rounding the real number upwards; n={24,14} represents that the value of n is 24 or 14.
[0086] In an optional embodiment, the historical load data is represented by X(t), and the historical load data X(t) is subjected to Fourier series expansion to obtain the following specific expression of X(t):
[0087]
[0088] Among them, X(t) is the historical load data, α0 is the DC component, and a i and b i are the sine and cosine coefficients, ω i is the angular frequency, N is the total number of load data;
[0089] The historical load data X(t) is reconstructed to obtain another expression of the historical load data X(t):
[0090] X(t)=α0+D(t)+W(t)+M(t),
[0091] Among them, α0+D(t) is the daily cycle component of historical load, W(t) is the weekly cycle component of historical load, and M(t) is the monthly cycle component of historical load;
[0092] It should be noted that, since the present invention is to explore the relationship between the power load data and the vibration noise data of the power equipment load under the rated working state, therefore, when reconstructing the historical load data, this embodiment ignores the high-frequency component and the low-frequency component obtained by the reconstruction. Since the load data of the power equipment under the rated working state is closely related to the date, for example, the residential electricity consumption shows the characteristics of low power load demand during the day and high power load demand at night, low power load demand on weekdays, low power load demand on weekends, low power load demand in summer and winter, and high power load demand in spring and autumn, the present invention decomposes the historical load data into the historical load daily cycle component, the historical load weekly cycle component and the historical load monthly cycle component.
[0093] In an optional embodiment, the historical load daily cycle component α0+D(t) is sampled once every hour and 24 times a day, that is, the period of D(t) is 24. In this embodiment, the historical load daily cycle component with a period of one day can be expressed as follows:
[0094]
[0095] {i|mod(i,12)=0,i=0,1,…,N-1}U{0},
[0096] Among them, Ω day is the angular frequency of the daily component of the historical load, and mod is the remainder operation;
[0097] The historical load cycle component W(t), where the sample is taken 168 times per week, that is, the period of W(t) is 168, the angular frequency of the historical load cycle component W(t) is expressed as follows:
[0098]
[0099] {i|mod(i,7)=0∩mod(i,12)≠0,i=0,1,…,N-1},
[0100] Among them, Ω week is the angular frequency of the historical load cycle component, mod is the remainder operation;
[0101] The historical load monthly cycle component M(t), wherein the month is calculated as 28 days and 4 weeks, wherein each month is sampled 672 times, and the angular frequency of the historical load monthly cycle component with a cycle of 28 days in this embodiment can be expressed as follows:
[0102]
[0103] {i|mod(i,28)=0∩mod(i,12)≠0,i=0,1,…,N-1∩mod(i,7)≠0},
[0104] Among them, Ω moon is the angular frequency of the monthly component of the historical load, and mod is the remainder operation.
[0105] In an optional embodiment, the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component are weighted and counted according to a preset weighted average algorithm to obtain historical daily load data, specifically including:
[0106]
[0107] Among them, D l (t) is the historical daily load data, T d is the minimum period of the daily component of the historical load, T w T d is the minimum period of the historical load cycle component, T m T d It is the minimum period of the monthly component of historical load.
[0108] Step S102: constructing historical load time series data based on a preset clustering algorithm and combining the historical daily load data.
[0109] In this embodiment, the historical load time series data is constructed based on the preset clustering algorithm and combined with the historical daily load data, specifically including:
[0110] The historical daily load data is converted into historical daily load matrix data, and a kernel principal component analysis algorithm is used to reduce the dimension of the high-dimensional historical daily load matrix data to obtain first historical daily load matrix data;
[0111] Transform the load matrix data of the first historical day to obtain the historical daily load curve;
[0112] According to a preset clustering algorithm, the optimal clustering result is obtained in combination with the historical daily load curve;
[0113] According to the optimal clustering result, historical load time series data is constructed.
[0114] By converting the historical daily load data into historical daily load matrix data and performing dimensionality reduction processing, and then using a clustering algorithm, the load can be classified according to date characteristics, thereby obtaining a more accurate relationship between load and time.
[0115] In an optional embodiment, the converting the historical daily load data into historical daily load matrix data, and using a kernel principal component analysis algorithm to reduce the dimension of the high-dimensional historical daily load matrix data to obtain the first historical daily load matrix data specifically includes:
[0116] The historical daily load data is converted into the following historical daily load matrix data X:
[0117]
[0118] Among them, t1, t2, ..., t m are different sampling times, l1,l2,…,l n The power load value sampled at each sampling time;
[0119] The Kernel Principal Component Analysis (KPCA) algorithm projects the data in the original space into a high-dimensional feature space through a nonlinear mapping, and then performs a PCA-based data processing algorithm in the high-dimensional feature space.
[0120] In an optional embodiment, the preset clustering algorithm is combined with the historical daily load curve to obtain the optimal clustering result, and the historical load time series data is constructed according to the optimal clustering result. The dates are clustered according to the historical daily load curve and the size of the power load value to obtain several clusters; the center point of each cluster is selected as the representative of each cluster in turn, and the dates and power load values corresponding to all the center points are simulated to obtain the historical load time series data.
[0121] Step S103: Acquire the sound pressure information of the electric equipment according to a preset acquisition method to obtain a standard sound pressure signal.
[0122] In this embodiment, the sound pressure information of the electric equipment is collected according to a preset collection method to obtain a standard sound pressure signal, which specifically includes:
[0123] The sound field is constructed with the power equipment as the center point, and the power equipment is decomposed into a plurality of collection surfaces; each collection surface is decomposed into a plurality of collection points, and the sound pressure signal of each collection point and collection surface is collected, and the total sound pressure signal of the power equipment is constructed according to the sound pressure signal of each collection point and collection surface:
[0124]
[0125] p ij (r)=ρ0ω∫ D q(r0)G(r,r0),
[0126]
[0127]
[0128] Among them, p i is the sound pressure signal of the i-th collection surface, A i (ω) is the acoustic transfer vector of the i-th acquisition surface, N is the number of acquisition points in the i-th acquisition surface; p ij (r) is the acoustic pressure signal of the jth acquisition point in the i-th acquisition surface; ρ0 is the average density of the medium; ω is the vibration angular frequency of the power equipment; D is the boundary of the sound field; q(r0) is the pressure generated by the vibration source of the power equipment, G(r,r0) is the Green function; r0 is the center point of the vibration source of the power equipment; r is the distance between the acquisition point j and the center point of the vibration source of the power equipment; m is the number of acquisition surfaces, v i (ω) is the vibration rate of the power equipment at the i-th collection surface; P is the sound pressure signal of the power equipment;
[0129] The sound pressure signal of the electric equipment is normalized to obtain a standard sound pressure signal of the electric equipment.
[0130] By constructing the sound field of the power equipment and decomposing it into multiple collection surfaces and collection points, a more comprehensive and accurate sound pressure signal can be collected, thereby obtaining a more accurate sound pressure signal of the power equipment. Finally, through normalization processing, large differences between signals of different levels of sound pressure signals can be avoided, thereby ensuring that each sound pressure signal collected is used for the same sound pressure level.
[0131] In this embodiment, the normalizing the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment specifically includes:
[0132] Obtain a preset normalization ratio, and use the DBFS normalization method to normalize the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment:
[0133] DBFS max =20log 10 (max|soundsignal old |),
[0134] DBFS min = = 20log 10(max|soundsignal old |),
[0135]
[0136] Among them, DBFS new is the maximum normalized ratio, DBFS min is the minimum normalized ratio, soundsignal old is the sound pressure signal of the power equipment before normalization, soundsignal new is the standard sound pressure signal of the power equipment obtained after normalization.
[0137] By normalizing, large differences between signals of different sound pressure levels can be avoided, thereby ensuring that each sound pressure signal collected is used for the same sound pressure level.
[0138] Step S104: extracting features from the standard sound pressure signal to obtain first noise data.
[0139] In this embodiment, the feature extraction of the standard sound pressure signal to obtain the first noise data specifically includes:
[0140] Performing signal enhancement on the standard sound pressure signal according to a time shift operation method to obtain a first standard sound pressure signal;
[0141] Performing time-frequency analysis on the first standard sound pressure signal according to a short-time Fourier transform method to obtain a first time-frequency spectrum;
[0142] According to a preset feature extraction method, feature extraction is performed on the first time-frequency spectrum to obtain corresponding amplitude spectrum features and phase spectrum features;
[0143] Convolutionally encoding the amplitude spectrum features and the phase spectrum features to obtain audio features in a high-dimensional feature space;
[0144] Extracting audio features in the high-dimensional feature space according to a preset time-frequency convolution method to obtain time-frequency features;
[0145] Performing phase decoding on the time-frequency feature to obtain a corresponding phase feature;
[0146] According to a preset feature screening method, the time-frequency features and phase features are screened to obtain an audio feature set;
[0147] According to the inverse Fourier transform method, initial noise data is obtained in combination with the audio feature set;
[0148] The initial noise data is filtered according to a low-pass filtering method to obtain first noise data.
[0149] The sound pressure signal is converted into a time-frequency spectrum, and convolution coding, phase decoding, etc. are performed, and finally feature screening and low-pass filtering are performed to obtain accurate low-frequency noise data.
[0150] In an optional embodiment, filtering the initial noise data according to a low-pass filtering method to obtain the first noise data refers to filtering the initial noise data using a low-pass filter, thereby filtering out high-frequency noise data to obtain low-frequency first noise data.
[0151] Step S105: inputting the historical load time series data and the first noise data into a preset initial evaluation model, and obtaining an optimal evaluation model according to a preset evaluation training method.
[0152] In this embodiment, the inputting of the historical load time series data and the first noise data into a preset initial evaluation model and obtaining the optimal evaluation model according to a preset evaluation training method specifically includes:
[0153] Inputting the historical load time series data and the first noise data into a preset convolutional neural network model, so that the convolutional neural network model can fit the historical load time series data and the first noise data in the time dimension;
[0154] A weighted error evaluation method is set in the output layer of the convolutional neural network model to enable the evaluation result of the convolutional neural network model output for the noise data;
[0155] According to the evaluation results, the convolutional neural network model is iteratively trained, and the evaluation results are updated until the evaluation results meet the preset standards, the training of the convolutional neural network model is completed, and the optimal evaluation model is obtained.
[0156] By inputting the historical load time series data and the first noise data into the preset convolutional neural network model and adding the A-weighted error evaluation method to the convolutional neural network model, the relationship between the load data and the noise data in the time series can be fully explored and the noise level can also be evaluated.
[0157] In an optional embodiment, the convolutional neural network model includes an input layer, a CNN layer, an LSTM layer and an output layer; the input layer is used to decompose and extract features of the historical load time series data and the first noise data, and the CNN layer includes three layers of one-dimensional convolution and maximum pooling; the LSTM layer is used to obtain the relationship between the features of the historical load time series data and the first noise data; the A-weighted error evaluation method of the output layer is used to evaluate the level of noise.
[0158] Step S106: acquiring daily load data of the power equipment to be evaluated, and inputting the data into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
[0159] This embodiment decomposes the historical load data of the power equipment into historical daily load data that is more closely related to the daily operation of the power equipment, and constructs the corresponding time series data based on this; by collecting the sound pressure information of the power equipment and performing feature extraction, accurate low-frequency noise data is obtained. Through the evaluation model, the relationship between load and noise is mined from three dimensions: load, noise, and time, so that the evaluation of low-frequency noise data can be obtained based on the daily load data that is most closely related to the power equipment, so that the evaluation results are consistent with the actual operation of the power equipment, the accuracy of the noise evaluation is improved, and a basis is provided for the noise reduction of the power equipment, thereby optimizing the construction of the power equipment.
[0160] Embodiment 2
[0161] Please refer to Figure 2 , which is a schematic diagram of the structure of a vibration noise evaluation device for electric power equipment provided by an embodiment of the present invention, comprising: a load data acquisition module 201, a load time series data acquisition module 202, a sound pressure signal acquisition module 203, a noise data acquisition module 204, an evaluation model acquisition module 205 and a vibration noise evaluation module 206;
[0162] The load data acquisition module 201 is used to acquire and decompose the historical load data of the power equipment to obtain the historical daily load data.
[0163] In this embodiment, the load data acquisition module 201 includes: a load data acquisition unit;
[0164] The load data acquisition unit is used to acquire historical load data of the power equipment, and perform data cleaning on the historical load data to obtain first historical load data;
[0165] According to the preset date feature type, combined with the Fourier decomposition method, the first historical load data is decomposed to obtain the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component;
[0166] According to a preset weighted average algorithm, the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component are weightedly counted to obtain historical daily load data.
[0167] The load time series data acquisition module 202 is used to construct historical load time series data based on a preset clustering algorithm in combination with the historical daily load data.
[0168] In this embodiment, the load time series data acquisition module 202 includes: a load time series data acquisition unit;
[0169] The load time series data acquisition unit is used to convert the historical daily load data into historical daily load matrix data, and use a kernel principal component analysis algorithm to reduce the dimension of the high-dimensional historical daily load matrix data to obtain first historical daily load matrix data;
[0170] Transform the load matrix data of the first historical day to obtain the historical daily load curve;
[0171] According to a preset clustering algorithm, the optimal clustering result is obtained in combination with the historical daily load curve;
[0172] According to the optimal clustering result, historical load time series data is constructed.
[0173] The sound pressure signal acquisition module 203 is used to collect the sound pressure information of the electric equipment according to a preset collection method to obtain a standard sound pressure signal.
[0174] In this embodiment, the sound pressure signal acquisition module 203 includes: a sound pressure signal acquisition unit;
[0175] The sound pressure signal acquisition unit is used to construct a sound field with the power equipment as the center point, decompose the power equipment into a plurality of collection surfaces; decompose each collection surface into a plurality of collection points, collect the sound pressure signal of each collection point and collection surface, and construct a total sound pressure signal of the power equipment according to the sound pressure signal of each collection point and collection surface:
[0176]
[0177] p ij (r)=ρ0ω∫ D q(r0)G(r,r0),
[0178]
[0179]
[0180] Among them, p i is the sound pressure signal of the i-th collection surface, A i (ω) is the acoustic transfer vector of the i-th acquisition surface, N is the number of acquisition points in the i-th acquisition surface; p ij(r) is the acoustic pressure signal of the jth acquisition point in the i-th acquisition surface; ρ0 is the average density of the medium; ω is the vibration angular frequency of the power equipment; D is the boundary of the sound field; q(r0) is the pressure generated by the vibration source of the power equipment, G(r,r0) is the Green function; r0 is the center point of the vibration source of the power equipment; r is the distance between the acquisition point j and the center point of the vibration source of the power equipment; m is the number of acquisition surfaces, v i (ω) is the vibration rate of the power equipment at the i-th collection surface; P is the sound pressure signal of the power equipment;
[0181] The sound pressure signal of the electric equipment is normalized to obtain a standard sound pressure signal of the electric equipment.
[0182] In this embodiment, the sound pressure signal acquisition module 203 includes: a normalization processing unit;
[0183] The normalization processing unit is used to obtain a preset normalization ratio, and use the DBFS normalization method to normalize the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment:
[0184] DBFS max =20log 10 (max|soundsignal old |),
[0185] DBFS min = = 20log 10 (max|soundsignal old |),
[0186]
[0187] Among them, DBFS new is the maximum normalized ratio, DBFS min is the minimum normalized ratio, soundsignal old is the sound pressure signal of the power equipment before normalization, soundsignal new is the standard sound pressure signal of the power equipment obtained after normalization.
[0188] The noise data acquisition module 204 is used to extract features from the standard sound pressure signal to obtain first noise data.
[0189] In this embodiment, the noise data acquisition module 204 includes: a noise data acquisition unit;
[0190] The noise data acquisition unit is used to perform signal enhancement on the standard sound pressure signal according to the time shift operation method to obtain a first standard sound pressure signal;
[0191] Performing time-frequency analysis on the first standard sound pressure signal according to a short-time Fourier transform method to obtain a first time-frequency spectrum;
[0192] According to a preset feature extraction method, feature extraction is performed on the first time-frequency spectrum to obtain corresponding amplitude spectrum features and phase spectrum features;
[0193] Convolutionally encoding the amplitude spectrum features and the phase spectrum features to obtain audio features in a high-dimensional feature space;
[0194] Extracting audio features in the high-dimensional feature space according to a preset time-frequency convolution method to obtain time-frequency features;
[0195] Performing phase decoding on the time-frequency feature to obtain a corresponding phase feature;
[0196] According to a preset feature screening method, the time-frequency features and phase features are screened to obtain an audio feature set;
[0197] According to the inverse Fourier transform method, initial noise data is obtained in combination with the audio feature set;
[0198] The initial noise data is filtered according to a low-pass filtering method to obtain first noise data.
[0199] The evaluation model acquisition module 205 is used to input the historical load time series data and the first noise data into a preset initial evaluation model, and obtain an optimal evaluation model according to a preset evaluation training method.
[0200] In this embodiment, the evaluation model acquisition module 205 includes: an evaluation model acquisition unit;
[0201] The evaluation model acquisition unit is used to input the historical load time series data and the first noise data into a preset convolutional neural network model, so that the convolutional neural network model can fit the historical load time series data and the first noise data in the time dimension;
[0202] A weighted error evaluation method is set in the output layer of the convolutional neural network model to enable the evaluation result of the convolutional neural network model output for the noise data;
[0203] According to the evaluation results, the convolutional neural network model is iteratively trained, and the evaluation results are updated until the evaluation results meet the preset standards, the training of the convolutional neural network model is completed, and the optimal evaluation model is obtained.
[0204] The vibration and noise evaluation module 206 is used to obtain the daily load data of the power equipment to be evaluated, and input it into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
[0205] This embodiment decomposes the historical load data of the power equipment into historical daily load data that is more closely related to the daily operation of the power equipment, and constructs the corresponding time series data based on this; by collecting the sound pressure information of the power equipment and performing feature extraction, accurate low-frequency noise data is obtained. Through the evaluation model, the relationship between load and noise is mined from three dimensions: load, noise, and time, so that the evaluation of low-frequency noise data can be obtained based on the daily load data that is most closely related to the power equipment, so that the evaluation results are consistent with the actual operation of the power equipment, the accuracy of the noise evaluation is improved, and a basis is provided for the noise reduction of the power equipment, thereby optimizing the construction of the power equipment.
[0206] Embodiment 3
[0207] The embodiments of the present invention provide a terminal device and a computer-readable storage medium.
[0208] The terminal device includes: one or more processors; a memory coupled to the processor and used to store one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a vibration noise evaluation method for an electric power equipment as described in the above embodiment.
[0209] The computer-readable storage medium stores a computer program thereon, wherein the computer program is executed by a processor to implement a vibration noise evaluation method for electric power equipment as described in the above embodiment.
[0210] In summary, the embodiment of the present invention decomposes the historical load data of the power equipment into historical daily load data that is more closely related to the daily operation of the power equipment, and constructs the corresponding time series data based on this; by collecting the sound pressure information of the power equipment and performing feature extraction, accurate low-frequency noise data is obtained. Through the evaluation model, the relationship between load and noise is mined from the three dimensions of load, noise and time, so that the evaluation of low-frequency noise data can be obtained based on the daily load data that is most closely related to the power equipment, so that the evaluation result is consistent with the actual operation of the power equipment, the accuracy of the noise evaluation is improved, and a basis is provided for the noise reduction of the power equipment, thereby optimizing the construction of the power equipment.
[0211] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims
1. A vibration noise evaluation method for electric power equipment, characterized in that: include: Obtain historical load data of power equipment and decompose it to obtain historical daily load data; According to a preset clustering algorithm, historical load time series data is constructed in combination with the historical daily load data; According to a preset collection method, the sound pressure information of the electric equipment is collected to obtain a standard sound pressure signal; Extracting features of the standard sound pressure signal to obtain first noise data; Inputting the historical load time series data and the first noise data into a preset initial evaluation model, and obtaining an optimal evaluation model according to a preset evaluation training method; The daily load data of the power equipment to be evaluated is obtained and input into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
2. A vibration noise evaluation method for electric power equipment as claimed in claim 1, characterized in that: The historical load data of the power equipment is decomposed to obtain the historical daily load data, which specifically includes: Acquire historical load data of the power equipment, and perform data cleaning on the historical load data to obtain first historical load data; According to the preset date feature type, combined with the Fourier decomposition method, the first historical load data is decomposed to obtain the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component; According to a preset weighted average algorithm, the initial historical load daily cycle component, the initial historical load weekly cycle component and the initial historical load monthly cycle component are weightedly counted to obtain historical daily load data.
3. A vibration noise evaluation method for electric power equipment according to claim 1, characterized in that: The method of constructing the historical load time series data according to the preset clustering algorithm and combining the historical daily load data specifically includes: The historical daily load data is converted into historical daily load matrix data, and a kernel principal component analysis algorithm is used to reduce the dimension of the high-dimensional historical daily load matrix data to obtain first historical daily load matrix data; Transform the load matrix data of the first historical day to obtain the historical daily load curve; According to a preset clustering algorithm, the optimal clustering result is obtained in combination with the historical daily load curve; According to the optimal clustering result, historical load time series data is constructed.
4. A vibration noise evaluation method for electric power equipment according to claim 1, characterized in that: The method of collecting the sound pressure information of the electric equipment according to the preset collection method to obtain the standard sound pressure signal specifically includes: The sound field is constructed with the power equipment as the center point, and the power equipment is decomposed into a plurality of collection surfaces; each collection surface is decomposed into a plurality of collection points, and the sound pressure signal of each collection point and collection surface is collected, and the total sound pressure signal of the power equipment is constructed according to the sound pressure signal of each collection point and collection surface: p ij (r)=ρ0ω∫ D q(r0)G(r,r0), Among them, p i is the sound pressure signal of the ith collection surface, A i (ω) is the acoustic transfer vector of the i-th acquisition surface, N is the number of acquisition points in the i-th acquisition surface; p ij (r) is the acoustic pressure signal of the jth acquisition point in the i-th acquisition surface; ρ0 is the average density of the medium; ω is the vibration angular frequency of the power equipment; D is the boundary of the sound field; q(r0) is the pressure generated by the vibration source of the power equipment, G(r,r0) is the Green function; r0 is the center point of the vibration source of the power equipment; r is the distance between the acquisition point j and the center point of the vibration source of the power equipment; m is the number of acquisition surfaces, v i (ω) is the vibration rate of the power equipment at the i-th collection surface; P is the sound pressure signal of the power equipment; The sound pressure signal of the electric equipment is normalized to obtain a standard sound pressure signal of the electric equipment.
5. A vibration noise evaluation method for electric power equipment as claimed in claim 4, characterized in that: The normalizing the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment specifically includes: Obtain a preset normalization ratio, and use the DBFS normalization method to normalize the sound pressure signal of the electric equipment to obtain a standard sound pressure signal of the electric equipment: DBFS max =20log 10 (max|soundsignal old |), DBFS min ==20log 10 (max|soundsignal old |), Among them, DBFS new is the maximum normalized ratio, DBFS min is the minimum normalized ratio, soundsignal old is the sound pressure signal of the power equipment before normalization, soundsignal new is the standard sound pressure signal of the power equipment obtained after normalization.
6. A vibration noise evaluation method for electric power equipment according to claim 1, characterized in that: The feature extraction of the standard sound pressure signal to obtain the first noise data specifically includes: Performing signal enhancement on the standard sound pressure signal according to a time shift operation method to obtain a first standard sound pressure signal; Performing time-frequency analysis on the first standard sound pressure signal according to a short-time Fourier transform method to obtain a first time-frequency spectrum; According to a preset feature extraction method, feature extraction is performed on the first time-frequency spectrum to obtain corresponding amplitude spectrum features and phase spectrum features; Convolutionally encoding the amplitude spectrum features and the phase spectrum features to obtain audio features in a high-dimensional feature space; Extracting audio features in the high-dimensional feature space according to a preset time-frequency convolution method to obtain time-frequency features; Performing phase decoding on the time-frequency feature to obtain a corresponding phase feature; According to a preset feature screening method, the time-frequency features and phase features are screened to obtain an audio feature set; According to the inverse Fourier transform method, initial noise data is obtained in combination with the audio feature set; The initial noise data is filtered according to a low-pass filtering method to obtain first noise data.
7. A vibration noise evaluation method for electric power equipment according to claim 1, characterized in that: The step of inputting the historical load time series data and the first noise data into a preset initial evaluation model and obtaining an optimal evaluation model according to a preset evaluation training method specifically includes: Inputting the historical load time series data and the first noise data into a preset convolutional neural network model, so that the convolutional neural network model can fit the historical load time series data and the first noise data in the time dimension; A weighted error evaluation method is set in the output layer of the convolutional neural network model to enable the evaluation result of the convolutional neural network model output for the noise data; According to the evaluation results, the convolutional neural network model is iteratively trained, and the evaluation results are updated until the evaluation results meet the preset standards, the training of the convolutional neural network model is completed, and the optimal evaluation model is obtained.
8. A vibration noise evaluation device for electric power equipment, characterized in that: include: A load data acquisition module, a load time series data acquisition module, a sound pressure signal acquisition module, a noise data acquisition module, an evaluation model acquisition module and a vibration noise evaluation module; The load data acquisition module is used to acquire historical load data of power equipment and decompose it to obtain historical daily load data; The load time series data acquisition module is used to construct historical load time series data based on a preset clustering algorithm and in combination with the historical daily load data; The sound pressure signal acquisition module is used to collect the sound pressure information of the electric equipment according to a preset collection method to obtain a standard sound pressure signal; The noise data acquisition module is used to extract features from the standard sound pressure signal to obtain first noise data; The evaluation model acquisition module is used to input the historical load time series data and the first noise data into a preset initial evaluation model, and obtain an optimal evaluation model according to a preset evaluation training method; The vibration and noise evaluation module is used to obtain the daily load data of the power equipment to be evaluated, and input it into the optimal evaluation model to obtain the vibration and noise evaluation result of the power equipment.
9. A terminal device, characterized in that: include: one or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the vibration noise evaluation method for electric power equipment as described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement a vibration noise evaluation method for electric power equipment according to any one of claims 1 to 7.
