CN114066214A - Power quality analysis method based on multi-fusion convolutional neural network - Google Patents
Power quality analysis method based on multi-fusion convolutional neural network Download PDFInfo
- Publication number
- CN114066214A CN114066214A CN202111336232.0A CN202111336232A CN114066214A CN 114066214 A CN114066214 A CN 114066214A CN 202111336232 A CN202111336232 A CN 202111336232A CN 114066214 A CN114066214 A CN 114066214A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- electric energy
- neural network
- convolutional neural
- 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.)
- Pending
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 29
- 238000004458 analytical method Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000000034 method Methods 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000012795 verification Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000010606 normalization Methods 0.000 claims description 16
- 239000002131 composite material Substances 0.000 claims description 11
- 238000013528 artificial neural network Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 7
- 230000001052 transient effect Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 9
- 238000004590 computer program Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000004927 fusion Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000005067 remediation Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
- G06F17/142—Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
- Y02P90/82—Energy audits or management systems therefor
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Human Resources & Organizations (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Pure & Applied Mathematics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Entrepreneurship & Innovation (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Biophysics (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Marketing (AREA)
- Evolutionary Biology (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Databases & Information Systems (AREA)
Abstract
The invention discloses a power quality analysis method based on a multi-fusion convolutional neural network, which comprises the steps of obtaining power signal data with superposition disturbance; preprocessing the electric energy signal data to remove noise; classifying the electric energy signal data after the noise is removed; constructing a multi-fusion convolutional neural network model; taking the classified electric energy signal as a data set, and dividing the data set into three parts including a training set for training data, a test set for testing data and a verification set for verifying data; selecting an electric energy signal with a fixed length and in a data set as input data, wherein a training target is a marked signal type; testing the error of the trained model, adjusting the model parameters according to the error, and repeating the relevant steps until the model is trained to be convergent; and outputting the model with the minimum error. Compared with a general convolutional neural network model, the method has the advantages of higher anti-noise effectiveness, higher training speed and better accuracy.
Description
Technical Field
The invention relates to the field of power quality analysis of power systems, in particular to a power quality analysis method based on a multi-fusion convolutional neural network
Background
In recent years, smart grids have grown rapidly. However, the quality of the electric energy provided by the power system is always an important incentive to influence the normal operation of industrial equipment, and thus the production efficiency and economic benefit of enterprises. The ultimate goal of the power system is to provide a stable, clean, distortion-free grid signal. Therefore, power quality analysis has become the first step in power system pollution detection and remediation. The widespread use of various types of nonlinear loads makes the grid signals more and more complex. The power supply change-over switch, the charging pile, the photovoltaic grid connection and the like can inject interference signals into a power grid, so that the voltage, the current and the frequency of the power grid fluctuate. In addition, maintaining and managing the electrical equipment consumes a great deal of time and money. Common power quality events include steady state and transient types such as voltage transients, interruptions, harmonics, and transients. In addition, the range of the power quality interference signal is various, and a single interference signal and a mixed interference signal exist, so that the intelligent requirement on the identification method is high.
Over the past several decades, many researchers have investigated the problem of detecting and classifying disturbances in the quality of electrical energy. These methods generally include three steps: signal processing, feature design, and classifiers. Currently, signal processing methods are used to improve the time-frequency resolution of signals. However, it is difficult to adapt to the variation of the mixed perturbation signal for a single parameter.
In the characteristic design stage, different manual characteristics are mainly designed to represent the power quality disturbance, such as standard energy difference, maximum amplitude peak value and global disturbance ratio of the natural modal function. But in practice most of the features are set according to the effect of the classifier. However, the design of this function lacks reference, and information may be lost due to human factors. Many advanced classifiers have been designed to process hand-made features in the classification phase. For example, support vector machines, extreme learning machines, and random forests are the most common methods of interference classification. In recent years, new classification methods such as dynamic and detrending fluctuation analysis have been used to classify extracted features. These methods have achieved the intended purpose, but some problems have been overlooked.
The first problem is that a single type of interference results in high performance of the classifier. In practice, the actual power quality disturbances usually comprise superimposed disturbances, wherein two or even three disturbances need to be considered. Another problem is the simplification of the interference detection process. Recently, there have been attempts to analyze complex power quality disturbances using bayesian networks. However, bayesian networks require additional a priori information and therefore accuracy may be limited by other conditions. And with the rapid development of deep learning technology, the adoption of convolutional neural networks and long-short term memory networks for feature extraction and classification becomes popular. However, LSTM requires a large amount of computation and storage. In addition, a regular six-layer depth CNN is introduced, and the power quality disturbance is directly classified. But the network parameters are too many, easily resulting in overfitting. And does not fully take into account the specific working mechanisms inside the network. This is a problem in the art.
Disclosure of Invention
In order to solve the above problems, the present invention provides a power quality analysis method based on a multi-fusion convolutional neural network, which is characterized by comprising the following steps:
s1: acquiring electric energy signal data with superposition disturbance;
s2: preprocessing the electric energy signal data to remove noise;
s3: classifying the electric energy signal data after the noise is removed;
s4: constructing a multi-fusion convolutional neural network model;
s5: taking the electric energy signal after the classification in the step S3 as a data set, and dividing the data set into three parts, including a training set for training data, a test set for testing data, and a verification set for verifying data;
s6: selecting the electric energy signal with fixed length and in the data set as input data, wherein the training target is the signal type marked in the step S3;
s7: testing the error of the trained model, adjusting the model parameters according to the error, and repeating the step S6 until the model is trained to be convergent;
s8: and outputting the model with the minimum error.
Preferably, the pretreatment comprises: filling the median of the missing values and standardizing the data.
Preferably, the data normalization comprises:
and carrying out fast Fourier transform on the electric energy signal data, wherein the formula is as follows:
wherein, l is 0,1,2 … …, N-1, x (N) is power signal data, N is the length of x (N), j is an imaginary unit;
normalizing the electric energy signal data after the fast Fourier transform according to the mean value and the standard deviation, wherein the normalization processing formula is as follows:
wherein X is a random variable and is input electric energy signal data after fast Fourier transform, and X*Is a normalized random variable of x, μ is the sample mean, σ is the sample standard deviation;
and forming a vector sequence by the normalized electric energy signal data according to the time sequence of the fast Fourier transform.
Preferably, the disturbance types in step S3 include:
voltage flicker with voltage dip, voltage trap with voltage transient, voltage harmonic with voltage transient.
Preferably, the step S4 includes:
s401: combining standard convolution and cavity convolution to provide one-dimensional composite convolution neural network framework Cr；
S402: introducing a one-dimensional composite convolution neural network framework, combining original signal information and physical characteristics based on fast Fourier transform, and combining the two types of characteristics into one layer;
s403: the introduction of a batch normalization layer speeds up training and minimizes the model structure to reduce the number of parameters.
Preferably, the one-dimensional complex convolutional neural network framework CrComprises the following steps:
wherein the content of the first and second substances,
is the weight of the convolution of the r-th layer,
is a deviation term.
Preferably, the formula for introducing the accelerated training of the batch normalization layer is as follows:
wherein, gamma is an automatic adjustment parameter for preventing the original disturbance characteristic distribution from being damaged, and beta is a displacement parameter for preventing the original disturbance characteristic distribution from being damaged.
Preferably, the signal type training formula in step S6 is as follows:
wherein, FcIs the output of the fully connected layer, p is the proportion of each point of the output in the population,
the correct type of interference is ultimately determined by the maximum probability score p.
Preferably, the step S7 includes:
s701: selecting a model which is trained by the verification set;
s702: performing error calculation on the model in the step S701 through the test set and the verification set;
s703: adjusting the model parameter gamma through an error result, and then repeating the step S6 until the error of the model reaches a preset requirement;
s704: the error reaches the model training result of the preset requirement;
wherein the number of times the model converges that the loop of step S6 is required is at least 20.
Preferably, the method further comprises S9: when the model is run in the actual environment and the deviation occurs between the predicted value and the actual value, the latest data is added to the step S2, and the model is retrained.
Has the advantages that: the method firstly models the analysis problem of the industrial power quality into a multi-classification problem, and then provides a complex power quality interference detection framework based on a multi-fusion convolutional neural network. The contribution is focused on automatic extraction and fusion of features from multiple sources. Firstly, the invention introduces an information fusion structure which takes time domain and frequency domain information of an electric energy quality interference signal as input; in addition, on the basis of standard convolution and cavity convolution, one-dimensional composite convolution is provided so as to improve the diversity of network characteristics; then, to speed up the training and prevent overfitting, a batch normalization method is used to adjust the distribution of features. The effectiveness of the industrial power quality analysis technology based on the multi-fusion convolutional neural network is verified through experiments. Compared with a characteristic design method and a general convolutional neural network model, the anti-noise effectiveness, higher training speed and better accuracy of the method are verified by simulation under different noises and experiments based on a hardware platform.
Drawings
FIG. 1 is a flow chart of one embodiment of the present invention;
FIG. 2 is a diagram of a model framework in an embodiment.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It is noted that the terms first, second, third, etc. are used herein to describe various components or features, but these components or features are not limited by these terms. These terms are only used to distinguish one element or part from another element or part. Terms such as "first," "second," and other numerical terms when used herein do not imply a sequence or order unless clearly indicated by the context. For convenience of description, spatially relative terms such as "inner", "outer", "upper", "lower", "left", "right", "upper", "left", "right", and the like are used herein to describe the orientation relation of the components or parts in the present embodiment, but these spatially relative terms do not limit the orientation of the technical features in practical use.
The power quality analysis method based on the multi-fusion convolutional neural network is provided. By using the fast Fourier transform to perform feature fusion and using the batch processing normalization layer, the complexity of using a neural network is reduced, the model construction is accelerated, and the important problem that the relation between insufficient control precision and overfitting cannot be effectively solved.
As shown in fig. 1 to 2, a method for analyzing power quality based on a multi-fusion convolutional neural network includes the following steps:
s1: and acquiring data, namely acquiring electric energy signal data with superposition disturbance.
S2: and preprocessing the electric energy signal data to remove noise. The pretreatment comprises the following steps: filling the median of the missing values and standardizing the data.
Wherein the data normalization comprises:
and carrying out fast Fourier transform on the electric energy signal data, wherein the formula is as follows:
wherein, l is 0,1,2 … …, N-1, x (N) is power signal data, N is the length of x (N), j is an imaginary unit;
normalizing the electric energy signal data after the fast Fourier transform according to the mean value and the standard deviation, wherein the normalization processing formula is as follows:
wherein X is a random variable and is input electric energy signal data after fast Fourier transform, and X*Is a normalized random variable of x, μ is the sample mean, σ is the sample standard deviation;
and forming a vector sequence by the normalized electric energy signal data according to the time sequence of the fast Fourier transform.
S3: and classifying the electric energy signal data after the noise is removed.
The disturbance types in step S3 include: voltage flicker and voltage sag, voltage trap and voltage transient, and voltage harmonic and voltage transient, but are not limited to the above categories.
S4: and constructing a multi-fusion convolutional neural network model.
The method is characterized in that one-dimensional composite convolution is provided on the basis of standard convolution and cavity convolution, and in order to improve the diversity of the characteristics extracted by the model, the specific multi-fusion convolution neural network model comprises the following contents:
s401: combining standard convolution and cavity convolution to provide one-dimensional composite convolution neural network framework Cr：
The one-dimensional composite convolution neural network framework CrComprises the following steps:
wherein the content of the first and second substances,
is the weight of the convolution of the r-th layer,
in addition, the one-dimensional complex convolution is that the standard convolution and the cavity convolution are respectively convoluted and then summed, and the sum is compared with 0 to obtain the maximum value.
S402: introducing a one-dimensional composite convolution neural network framework, combining original signal information and physical characteristics based on fast Fourier transform, and combining the two types of characteristics into one layer; here, the original signal is the power signal information after the noise is removed in step S3.
S403: introducing a batch normalization layer to accelerate training and minimizing a model structure to reduce the number of parameters;
the formula for introducing the accelerated training of the batch normalization layer is as follows:
wherein, gamma is an automatic adjustment parameter for preventing the original disturbance characteristic distribution from being damaged, and beta is a displacement parameter for preventing the original disturbance characteristic distribution from being damaged.
S5: and taking the electric energy signal after the classification in the step S3 as a data set, and dividing the data set into three parts including a training set for training data, a test set for testing data, and a verification set for verifying data.
S6: selecting a fixed-length electric energy signal in a data set as input data, wherein the training target is the signal type marked in the step S3.
The explanation here for a fixed length comprising the original signal and the fast fourier transform vector is:
the fixed length is the input length of the model, the original signal 640 and the frequency domain signal after FFT (fast fourier transform) are 320, the voltage standard is 50hz, and the sampling frequency of the wave recorder is 3200hz, so that voltage information of 10 periods is taken from 640 points.
Because of the symmetry of the FFT result, 1/2 is taken, i.e., 320, and data with a length of 320 is input.
Selecting a fixed-length vector sequence comprising an original signal and a fast Fourier transform as model input, wherein a training target is a signal type marked in a preprocessing process, namely, data of a classified signal type is input into the model, and the model is trained to carry out type, wherein one part is used for training, and the other part is used for detecting the accuracy of the model.
The function of the training model classification is softmax, and the signal type training formula is as follows:
wherein, FcIs made ofThe output of the connected layer, p is the proportion of each point of the output in the population,
the correct type of interference is ultimately determined by the maximum probability score p.
S7: and (4) testing the error of the trained model, adjusting the model parameters according to the error, and repeating the step S6 until the trained model converges, wherein the convergence is that the trained model is matched with the classification without the step S3 after the training set is trained by the model. The number of times the model converges the loop of step S6 is at least 20.
The step S7 includes:
s701: selecting a model which is trained by the verification set;
s702: performing error calculation on the model in the step S701 through the test set and the verification set;
s703: adjusting the model parameter gamma through an error result, and then repeating the step S6 until the error of the model reaches a preset requirement;
s704: and the error reaches the model training result required by the preset.
S8: and outputting the model with the minimum error.
S9: when the model is run in the actual environment and the deviation occurs between the predicted value and the actual value, the latest data is added to the step S2, and the model is retrained.
The principle of the method is as follows:
data acquisition and data preprocessing:
preprocessing the data, including but not limited to: filling the median of the missing values, standardizing data and the like; calculating the fast Fourier transform of the preprocessed signal, and normalizing the fast Fourier transform according to the mean value and the standard deviation; forming a vector sequence by the processed signals and the fast Fourier transform according to the time sequence; and determining the disturbance type of the signal and classifying the signal.
Constructing a multi-fusion convolutional neural network model:
the method comprises the steps of providing one-dimensional composite convolution on the basis of standard convolution and cavity convolution in order to improve the diversity of features extracted by a model; a multi-fusion one-dimensional convolution neural network framework is introduced, original signal information and physical characteristics are combined together based on fast Fourier transform, the two types of characteristics are combined into one layer, the situation that a batch processing normalization layer is introduced to accelerate training due to complexity of a two-dimensional convolution neural network is avoided, and the number of parameters is reduced due to a minimized model structure.
The data set is divided into three parts: the training set is used for training data, the testing set is used for testing a training result, and the verification set is used for verifying a model; selecting a fixed-length vector sequence comprising an original signal and a fast Fourier transform vector as model input, wherein a training target is a signal type marked in a preprocessing process, and the model is trained for multiple times until convergence; verifying the trained model by using a verification set, adjusting model parameters by comparing the precision and the error of the test set and the verification set, and selecting the model with the best performance of the verification set as a training result by training for multiple times; and (4) running the model in the actual environment, and adding the latest data into the training set to train the model again when the predicted value and the actual value have larger deviation.
Aiming at the defects in the prior art, the invention aims to provide the electric energy quality analysis method based on the multi-fusion convolutional neural network, which improves the diversity of feature extraction by fusing the physical features extracted by fast Fourier transform, introducing the construction of a batch normalization layer acceleration model and applying a one-dimensional composite convolution method, so that the network has the features of avoiding the complexity of a two-dimensional convolutional neural network, and the accuracy of the electric energy quality analysis in the power system industry is effectively improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The present invention is not limited to the above embodiments, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are included in the scope of the claims of the present invention which are filed as the application.
Claims (10)
1. A power quality analysis method based on a multi-fusion convolutional neural network is characterized by comprising the following steps:
s1: acquiring electric energy signal data with superposition disturbance;
s2: preprocessing the electric energy signal data to remove noise;
s3: classifying the electric energy signal data after the noise is removed;
s4: constructing a multi-fusion convolutional neural network model;
s5: taking the electric energy signal after the classification in the step S3 as a data set, and dividing the data set into three parts, including a training set for training data, a test set for testing data, and a verification set for verifying data;
s6: selecting the electric energy signal with fixed length and in the data set as input data, wherein the training target is the signal type marked in the step S3;
s7: testing the error of the trained model, adjusting the model parameters according to the error, and repeating the step S6 until the model is trained to be convergent;
s8: and outputting the model with the minimum error.
2. The method according to claim 1, wherein the preprocessing comprises: filling the median of the missing values and standardizing the data.
3. The method according to claim 2, wherein the data normalization comprises:
and carrying out fast Fourier transform on the electric energy signal data, wherein the formula is as follows:
wherein, l is 0,1,2 … …, N-1, x (N) is power signal data, N is the length of x (N), j is an imaginary unit;
normalizing the electric energy signal data after the fast Fourier transform according to the mean value and the standard deviation, wherein the normalization processing formula is as follows:
wherein X is a random variable and is input electric energy signal data after fast Fourier transform, and X*Is a normalized random variable of x, μ is the sample mean, σ is the sample standard deviation;
and forming a vector sequence by the normalized electric energy signal data according to the time sequence of the fast Fourier transform.
4. The method for analyzing the quality of the electric energy based on the multi-fusion convolutional neural network as claimed in claim 2, wherein the disturbance type in the step S3 includes:
voltage flicker with voltage dip, voltage trap with voltage transient, voltage harmonic with voltage transient.
5. The method for analyzing the quality of the electric energy based on the multi-fusion convolutional neural network as claimed in claim 3, wherein the step S4 comprises:
s401: combining standard convolution and cavity convolution to provide one-dimensional composite convolution neural network framework Cr；
S402: introducing a one-dimensional composite convolution neural network framework, combining original signal information and physical characteristics based on fast Fourier transform, and combining the two types of characteristics into one layer;
s403: the introduction of a batch normalization layer speeds up training and minimizes the model structure to reduce the number of parameters.
6. The method according to claim 5, wherein the one-dimensional complex convolutional neural network framework C is a multi-fusion convolutional neural network-based power quality analysis methodrComprises the following steps:
wherein the content of the first and second substances,
is the weight of the convolution of the r-th layer,
is a deviation term.
7. The method according to claim 6, wherein the formula for introducing the accelerated training of the batch normalization layer is as follows:
wherein, gamma is an automatic adjustment parameter for preventing the original disturbance characteristic distribution from being damaged, and beta is a displacement parameter for preventing the original disturbance characteristic distribution from being damaged.
8. The method for analyzing the power quality based on the multi-fusion convolutional neural network of claim 1, wherein the signal type training formula in the step S6 is as follows:
wherein, FcIs the output of the fully connected layer, p is the proportion of each point of the output in the population,
the correct type of interference is ultimately determined by the maximum probability score p.
9. The method for analyzing the quality of the electric energy based on the multi-fusion convolutional neural network as claimed in claim 6, wherein the step S7 comprises:
s701: selecting a model which is trained by the verification set;
s702: performing error calculation on the model in the step S701 through the test set and the verification set;
s703: adjusting the model parameter gamma through an error result, and then repeating the step S6 until the error of the model reaches a preset requirement;
s704: the error reaches the model training result of the preset requirement;
wherein the number of times the model converges that the loop of step S6 is required is at least 20.
10. The method for analyzing the quality of the electric energy based on the multi-fusion convolutional neural network as claimed in claim 1, further comprising S9: when the model is run in the actual environment and the deviation occurs between the predicted value and the actual value, the latest data is added to the step S2, and the model is retrained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336232.0A CN114066214A (en) | 2021-11-12 | 2021-11-12 | Power quality analysis method based on multi-fusion convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111336232.0A CN114066214A (en) | 2021-11-12 | 2021-11-12 | Power quality analysis method based on multi-fusion convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114066214A true CN114066214A (en) | 2022-02-18 |
Family
ID=80275152
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111336232.0A Pending CN114066214A (en) | 2021-11-12 | 2021-11-12 | Power quality analysis method based on multi-fusion convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114066214A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116362300A (en) * | 2022-06-29 | 2023-06-30 | 国网河南省电力公司 | Regional power grid abnormal disturbance quantity prediction method, device, medium and electronic equipment |
-
2021
- 2021-11-12 CN CN202111336232.0A patent/CN114066214A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116362300A (en) * | 2022-06-29 | 2023-06-30 | 国网河南省电力公司 | Regional power grid abnormal disturbance quantity prediction method, device, medium and electronic equipment |
| CN116362300B (en) * | 2022-06-29 | 2024-02-09 | 国网河南省电力公司 | Regional power grid abnormal disturbance quantity prediction method, device, medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107462785B (en) | The more disturbing signal classifying identification methods of power quality based on GA-SVM | |
| Wang et al. | Decision tree based online stability assessment scheme for power systems with renewable generations | |
| CN111523785A (en) | Power system dynamic security assessment method based on generation countermeasure network | |
| CN108920863B (en) | Method for establishing energy consumption estimation model of robot servo system | |
| CN110994604B (en) | Power system transient stability assessment method based on LSTM-DNN model | |
| CN112508442B (en) | Transient stability assessment method and system based on automatic and interpretable machine learning | |
| CN110263839B (en) | Power system load static characteristic online intelligent identification method based on big data | |
| CN110718910A (en) | Transient stability evaluation method for Bayesian optimization LightGBM | |
| CN111523778A (en) | Power grid operation safety assessment method based on particle swarm algorithm and gradient lifting tree | |
| CN112200038B (en) | CNN-based quick identification method for oscillation type of power system | |
| CN109298225B (en) | Automatic identification model system and method for abnormal state of voltage measurement data | |
| CN106897945A (en) | The clustering method and equipment of wind power generating set | |
| CN112580588A (en) | Intelligent flutter signal identification method based on empirical mode decomposition | |
| CN112200694A (en) | Dominant instability mode identification model construction and application method based on graph neural network | |
| CN115166618A (en) | Current transformer error evaluation method for non-stable output | |
| CN112069723A (en) | Method and system for evaluating transient stability of power system | |
| CN110991689B (en) | Distributed photovoltaic power generation system short-term prediction method based on LSTM-Morlet model | |
| CN115526258A (en) | Power system transient stability evaluation method based on Spearman correlation coefficient feature extraction | |
| CN114066214A (en) | Power quality analysis method based on multi-fusion convolutional neural network | |
| CN113361737A (en) | Abnormity early warning method and system for photovoltaic module | |
| CN114583767A (en) | Data-driven wind power plant frequency modulation response characteristic modeling method and system | |
| CN117151488A (en) | Method, system, storage medium and equipment for expanding cold tide and strong wind weather sample | |
| CN110717623B (en) | Photovoltaic power generation power prediction method, device and equipment integrating multiple weather conditions | |
| CN111539508A (en) | Generator excitation system parameter identification algorithm based on improved wolf algorithm | |
| CN116305683A (en) | Power system transient stability evaluation method and system based on sample equalization |
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 |