A harmonic source modeling method based on time-frequency feature extraction

Abstract

The application provides a harmonic source modeling method based on time-frequency feature extraction, first determines harmonic principal component frequencies of a harmonic source; then extracts harmonic voltage amplitude and phase angle corresponding to the harmonic principal component frequencies of the harmonic source, obtains a frequency domain feature set of the harmonic source, reconstructs the frequency domain feature set through IDFT to form a harmonic source frequency domain feature component matrix, and splices the frequency domain feature component matrix with an original voltage sequence collected by a synchronous phasor measurement device to obtain a time-frequency feature component matrix; the component matrix is converted into a gray scale image by using a matrix normalization method, and is converted into a three-channel time-frequency feature component color image by using a pseudo-color coding technology; finally, the collected harmonic source voltage data and the time-frequency feature component color image are taken as inputs of the model, the collected corresponding harmonic source current data are taken as outputs of the model, the network model is continuously trained through the measured data until accurate mapping of a relationship between harmonic voltage and harmonic current of the harmonic source is realized, and a harmonic source data-driven model is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of electric power big data, harmonic source modeling, and the like, and particularly relates to a harmonic source modeling method based on time-frequency feature extraction. BACKGROUND

[0002] With the rapid improvement of energy structure transformation and industrialization level of national economy, the number of nonlinear loads such as electric arc furnaces, electric locomotives and power electronic devices has increased rapidly, and the harmonic pollution problem brought by it has become increasingly serious. Accurate understanding of the harmonic characteristics of harmonic source load is helpful for harmonic pollution prevention and improvement of power quality level. Therefore, it is of great significance to construct an accurate and reasonable harmonic source model.

[0003] At present, harmonic source modeling can be divided into frequency domain modeling and time domain modeling according to the form of the established harmonic voltage and harmonic current expression. The frequency domain harmonic source model can output the vector value of each harmonic emitted by the harmonic source, and can reflect the long-term steady-state frequency domain characteristics of the harmonic source. However, this type of model usually needs to update the model parameters when the operating state of the harmonic source changes, so it is often used for harmonic sources with stable working conditions, and is difficult to apply to harmonic sources with frequent working condition changes or strong time-varying characteristics. The time domain modeling reflects the time-varying characteristics of the harmonic source by introducing time sequence information. The obtained harmonic source voltage and current time domain waveform contains the instantaneous change characteristics of the harmonic source, and is often used in the establishment of accurate harmonic source models. However, this type of model is often difficult to guarantee the modeling accuracy and generalization for complex harmonic sources with unknown specific topology and unknown mechanism characteristics.

[0004] Under the background of various types of harmonic sources with different properties, the harmonic characteristics of harmonic sources are greatly different due to the influence of harmonic generation mechanism. If only the dynamic time-varying characteristics of the harmonic source are concerned, the complexity of the constructed harmonic source model will be too high, the generalization will be insufficient, and overfitting will easily occur. If only the frequency domain steady-state characteristics of the harmonic source are concerned, the harmonic characteristics of the harmonic source running in a non-stationary state cannot be accurately reflected, and the time-varying nature of the harmonic source cannot be characterized. In order to more comprehensively reflect the harmonic characteristics of the harmonic source and improve the accuracy of the harmonic source model, the short-term time-varying characteristics and long-term steady-state frequency domain characteristics of the harmonic source should be extracted for analysis and modeling. At present, some scholars add a constraint on the frequency domain to the minimum target error function of the RBF neural network, and improve the anti-interference ability of the model to time-varying noise by introducing frequency domain characteristics. However, the frequency domain error function in this method is complex and difficult to solve, and the corresponding parameters of the time domain and frequency domain error functions in the neuron weight update are not easy to determine. Therefore, how to consider the steady-state frequency domain characteristics and dynamic time-varying nature of the harmonic source to improve the accuracy and generalization of the current harmonic source modeling is a big problem in the current harmonic source modeling. SUMMARY

[0005] In view of the limitations of the prior art, with the wide application of synchronous phasor measurement devices, a large amount of accurate harmonic source measured data is continuously accumulated, and the time-domain modeling method of the harmonic source based on data driving has made certain progress. Such method starts from the external characteristics of the harmonic source, without knowing the specific topology, and only needs the measured data of the harmonic source port to establish the time-domain model of the harmonic source. Among them, deep learning can achieve good results in time-domain modeling of harmonic sources with strong nonlinearity and complex internal mechanism due to the powerful non-linear mapping capability of neural network. In view of the fact that the existing synchronous phasor measurement device can provide rich high-precision harmonic measurement data, the present application proposes a harmonic source data-driven modeling method based on time-frequency feature extraction.

[0006] The method uses a heterogeneous feature input combination reflecting the dynamic time-varying characteristics and frequency domain steady-state characteristics of the harmonic source based on time domain analysis method, and introduces a harmonic source voltage and current mapping model combined with multiple convolutional neural network (MCNN) and bidirectional long short term memory network (BiLSTM).

[0007] The design points of the scheme include:

[0008] (1) A time-frequency feature extraction method is proposed, which can accurately extract the key frequency domain features affecting the harmonic characteristics of the harmonic source by analyzing the harmonic principal components of the harmonic source, and enhance and fuse the time-frequency features by constructing a time-frequency feature component color map.

[0009] (2) An MCNN model is proposed, which extracts features from the time domain feature extraction module and the frequency domain feature extraction module, and realizes the simultaneous extraction of different feature information of multi-channel and multi-dimensional input through the parallel structure of the two paths.

[0010] (3) An MCNN-BiLSTM model is proposed, which adopts the serial connection of MCNN and BiLSTM, and inputs the time-frequency domain fusion features extracted by the MCNN model into the BiLSTM network to obtain the feature information of the harmonic voltage in the whole time domain.

[0011] The specific technical solutions are as follows:

[0012] A harmonic source modeling method based on time-frequency feature extraction:

[0013] Firstly, the harmonic source voltage and current data collected by the synchronous phasor measurement device are normalized and subjected to FFT analysis; the harmonic component content rate is calculated, and the harmonic principal component frequency of the harmonic source is determined through the set threshold judgment rule;

[0014] Then, the harmonic voltage amplitude and phase angle corresponding to the harmonic principal component frequency of the harmonic source are extracted to obtain a frequency domain feature set of the harmonic source; the frequency domain feature set is reconstructed by IDFT to form a harmonic source frequency domain feature component matrix, and the time-frequency feature component matrix is obtained by splicing the original voltage sequence collected by the synchronous phasor measurement device; the component matrix is converted into a grayscale image by using a matrix normalization method, and is converted into a three-channel time-frequency feature component color image by using a pseudo-color coding technology;

[0015] Finally, the harmonic source voltage data collected by the synchronous phasor measurement device and the time-frequency feature component color image are used as inputs of the MCNN-BiLSTM model, the corresponding harmonic source current data collected by the synchronous phasor measurement device are used as outputs of the MCNN-BiLSTM model, the network model is continuously trained by using the measured data until the accurate mapping of the relationship between the harmonic voltage and the harmonic current of the harmonic source is realized, and a harmonic source data-driven model is obtained.

[0016] The MCNN-BiLSTM model is a harmonic source voltage and current mapping model composed of a multiple convolutional neural network MCNN and a bidirectional long short-term memory network BiLSTM.

[0017] Further, in the MCNN-BiLSTM model, the MCNN model extracts features from the time domain feature extraction module and the frequency domain feature extraction module, and the two parallel structures simultaneously extract different feature information of the multi-channel and multi-dimensional inputs; the MCNN and the BiLSTM are connected in series, the time-frequency domain fusion features extracted by the MCNN model are input into the BiLSTM network to obtain the feature information of the harmonic voltage in the entire time domain.

[0018] Further, the specific steps of the harmonic principal component analysis of the harmonic source are as follows:

[0019] Firstly, the original data obtained by the synchronous phasor measurement device is normalized to unify the scale of the original measurement data of the harmonic source:

[0020]

[0021] In the formula: is the normalized voltage and current sequence, U and I are the original voltage and current sequences measured by the synchronous phasor measurement device, U max and I max are the extreme values of the original voltage and current sequences measured by the synchronous phasor measurement device;

[0022] Secondly, the normalized voltage and current sequence is subjected to fast Fourier transform to obtain the corresponding kth harmonic component:

[0023]

[0024] wherein: U k , is the kth harmonic voltage amplitude and phase angle; I k , is the kth harmonic current amplitude and phase angle;

[0025] Then, the harmonic voltage content ratio HRU k and the harmonic current content ratio HRI k are calculated respectively:

[0026]

[0027] wherein: U1, I1 are the fundamental voltage and fundamental current amplitudes respectively;

[0028] Finally, according to the following formula, the harmonic frequencies with harmonic ratio HRU k and HRI k greater than 1% are selected as the harmonic principal component frequencies, and a harmonic source harmonic principal component frequency set γ is formed;

[0029]

[0030] Further, according to the harmonic principal component frequency set, the corresponding kth harmonic voltage amplitude and phase angle are extracted, and a harmonic source frequency domain feature set F is obtained. The steps of converting F into a time-frequency feature component color map are as follows:

[0031] First, the inverse Fourier transform is used to reconstruct each harmonic voltage component:

[0032]

[0033] wherein: U k (t) represents the kth harmonic voltage feature component, and N is the number of sampling points;

[0034] Second, the harmonic voltage feature components corresponding to the harmonic principal components {U1(t), U3(t), …, U k (t)} obtained by the inverse Fourier transform are spliced with the original voltage sequence U(t) collected by the synchronous phasor measurement device, so as to obtain a time-frequency feature component matrix {U1(t), U3(t), …, U k (t), U(t)};

[0035] Then, the time-frequency feature component matrix is processed by the following formula, so that the data are all normalized to the range of (0, 255), and then the time-frequency feature is corresponded with the gray value to obtain a gray image of the time-frequency feature;

[0036]

[0037] wherein G(m, n) represents the value of the gray scale matrix in the mth row and nth column; x i represents the value of the ith data point; x min represents the minimum value and maximum value in the time-frequency feature component matrix, respectively; max

[0038] Finally, the gray scale image is converted into a three-channel RGB color image by using pseudo-color encoding to obtain a time-frequency feature component color image.

[0039] Further, the time domain feature extraction module adopts a double-layer 1D convolution network structure, each layer of the 1D convolution network comprising a 1D convolution layer, a 1D regularization layer and a 1D pooling layer; the harmonic source voltage data X1 collected by the synchronous phasor measurement device is taken as input and sequentially passes through the two layers of 1D convolution network to finally obtain a feature vector Y 1DCNN :

[0040] Y 1DCNN = F 1DCNN (X1) (19)

[0041] wherein F 1DCNN represents the time domain feature extraction module.

[0042] Further, the frequency domain feature learning module adopts a double-layer 2D convolution network, each layer of the 2D convolution network comprising a 2D convolution layer, a 2D regularization layer and a 2D pooling layer; the time-frequency feature component color image X2 is taken as input and sequentially passes through the two layers of 2D convolution network to finally obtain a feature vector Y 2DCNN :

[0043] Y 2DCNN = F 2DCNN (X2) (20)

[0044] wherein F 2DCNN represents the frequency domain feature extraction module.

[0045] Further, the feature vectors output by the time domain feature extraction module and the frequency domain feature extraction module are flattened and converted into two feature vectors of the same size by a full connection layer; the two feature vectors are fused by splicing to obtain a fused feature Y cat , which is a local deep representation of the harmonic source feature time-frequency feature, and the splicing operation is represented as follows:

[0046]

[0047] wherein Y cat represents the fused feature after splicing, flatten is a flattening function, and Fc is a full connection layer function.

[0048] ​Further, the input of the BiLSTM network is the fusion feature Y cat , the output of the model is the harmonic current X out , specifically:

[0049] X out =F BiLSTM (Y cat ) (22)

[0050] In the formula: F BiLSTM represents the BiLSTM network.

[0051] And an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program comprising a harmonic source data-driven model obtained by the above-mentioned harmonic source modeling method based on time-frequency feature extraction.

[0052] A non-transitory computer-readable storage medium having stored thereon a computer program, the computer program comprising a harmonic source data-driven model obtained by the above-mentioned harmonic source modeling method based on time-frequency feature extraction.

[0053] Compared with the prior art, the present application and its preferred schemes can fully exploit the advantages of high-quality harmonic data collection of the synchronous phasor measurement device, realize data-driven accurate modeling of the harmonic source model, and provide support for harmonic pollution prevention and improvement of power quality. The time-frequency feature extraction method proposed in the present application can accurately extract the harmonic principal component frequency that affects the harmonic characteristics of the harmonic source, and use IDFT and pseudo-color coding technology to perform feature dimensionality increase and fusion of the time-domain features on the discrete frequency-domain features. The obtained time-frequency feature color map has more rich spatiotemporal feature information and stronger feature expressiveness than the 1-dimensional time-domain waveform. The proposed MCNN model can fully exploit the correlation between the short-term local features in the time domain and the harmonic components of each frequency in the frequency domain by using different scale convolution blocks and different size convolution kernels. Compared with a single structure CNN model, the MCNN can extract more rich and comprehensive harmonic source feature information. The MCNN-BiLSTM model has better model accuracy in single harmonic source modeling and complex harmonic source network modeling, and can more comprehensively reflect the harmonic characteristics of the harmonic source. BRIEF DESCRIPTION OF DRAWINGS

[0054] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0055] Figure 1 The present application is an embodiment model construction process and network structure schematic diagram. DETAILED DESCRIPTION

[0056] In the following, specific embodiments of the present application will be described in detail with reference to the accompanying drawings. From these descriptions, those skilled in the art will be able to understand the present application and implement it. The features of the various embodiments can be combined to obtain new embodiments, or to replace certain features in certain embodiments, to obtain other preferred embodiments, without departing from the principles of the present application.

[0057] It is to be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, devices, components and / or combinations thereof, but do not preclude the presence or addition of one or more other features, steps, operations, devices, components and / or combinations thereof.

[0058] In order that the character and advantages of the patent will be more clearly understood, in the following, specific examples will be described in detail as follows: 1 Harmonic source harmonic principal component analysis

[0059] In order to more accurately extract the frequency domain features of the harmonic source, it is necessary to determine the harmonic principal component emitted by the harmonic source. The specific steps are as follows:

[0060] Firstly, the original data obtained by the synchronous phasor measurement device is normalized to unify the scale of the original measurement data of the harmonic source. The specific formula is as follows:

[0061]

[0062] In the formula: is the normalized voltage and current sequence, U and I are the original voltage and current sequence measured by the synchronous phasor measurement device, and U max , max are the extreme values of the original voltage and current sequence measured by the synchronous phasor measurement device.

[0063] Secondly, the normalized voltage and current sequence is subjected to fast Fourier transform (FFT) to obtain the corresponding kth harmonic component:

[0064]

[0065] In the formula: U k , is the kth harmonic voltage amplitude and phase angle; I k , is the kth harmonic current amplitude and phase angle.

[0066] Then, the harmonic voltage content rate HRU k is calculated respectively.Harmonic current inclusion rate HRI k :

[0067]

[0068] In the formula: U1, I1 are the fundamental voltage and fundamental current amplitude respectively.

[0069] Finally, all HRU is selected according to the following formula k And HRI k The harmonic frequency with a proportion greater than 1% is taken as the harmonic principal component frequency, and constitutes the harmonic source harmonic principal component frequency set γ.

[0070]

[0071] 2 Time-frequency characteristic component color chart

[0072] According to the harmonic principal component frequency set obtained in section 1, the corresponding k-th harmonic voltage amplitude and phase angle are extracted, and the harmonic source frequency domain feature set is obtained The steps of converting F into a time-frequency characteristic component color chart are as follows:

[0073] Firstly, the inverse discrete Fourier transform (IDFT) is used to reconstruct each harmonic voltage component, and the specific formula is as follows:

[0074]

[0075] In the formula: U k (t) represents the k-th harmonic voltage characteristic component, and N is the number of sampling points.

[0076] Secondly, the harmonic voltage characteristic components corresponding to the harmonic principal components obtained after the inverse Fourier transform {U1(t), U3(t), …, U k (t)} are spliced with the original voltage sequence U(t) collected by the synchronous phasor measurement device, so as to obtain the time-frequency characteristic component matrix {U1(t), U3(t), …, U k (t), U(t)}.

[0077] Then, the time-frequency characteristic component matrix is processed by the following formula, so that the data is normalized to the range of (0, 255), and then it is corresponding to the computer internal code gray value, to obtain the gray image of time-frequency characteristics.

[0078]

[0079] In the formula: G(m, n) represents the value of the gray matrix in the mth row and nth column; x i represents the value of the ith data point; x min and xmax These represent the minimum and maximum values ​​in the time-frequency characteristic component matrix, respectively.

[0080] Finally, the grayscale image is converted into a three-channel RGB color image using pseudo-color coding technology to obtain a time-frequency feature component color map.

[0081] 3MCNN-BiLSTM network model

[0082] 3.1MCNN network

[0083] Convolutional neural networks (CNNs) are deep learning algorithms whose convolutional layers can extract local features from input data through convolution operations, exhibiting good performance in processing multidimensional nonlinear data. Since the setting of the convolutional kernel greatly affects the quality of feature extraction, and given that the time-domain features of harmonic sources exhibit complex time-varying characteristics while their frequency-domain features are relatively stable, this invention proposes a time-domain feature extraction module and a frequency-domain feature extraction module to improve the effectiveness of feature extraction.

[0084] (1) Temporal feature extraction module

[0085] The temporal feature learning module adopts a two-layer 1D convolutional network (1DCNN) structure. Each layer of the 1DCNN contains a 1D convolutional layer, a 1D regularization layer, and a 1D pooling layer. The harmonic source voltage data X1 acquired by the synchronous phasor measurement device is used as input, and sequentially passed through the two layers of the 1DCNN to finally obtain the feature vector Y. 1DCNN Specifically:

[0086] Y 1DCNN =F 1DCNN (X1) (29)

[0087] In the formula: F 1DCNN This indicates the temporal feature extraction module.

[0088] (2) Frequency domain feature extraction module

[0089] The frequency domain feature learning module employs a two-layer 2D convolutional network (2DCNN), where each layer of the 2DCNN contains a 2D convolutional layer, a 2D regularization layer, and a 2D pooling layer. The time-frequency feature component color image X2 obtained in Section 2 is used as input and passed through the two layers of the 2DCNN to finally obtain the feature vector Y. 2DCNN Specifically:

[0090] Y 2DCNN =F 2DCNN (X2) (30)

[0091] In the formula: F 2DCNN This indicates the frequency domain feature extraction module.

[0092] 3.2 Feature fusion layer

[0093] After the feature vectors output by the time domain feature extraction module and the frequency domain feature extraction module are flattened, they are converted into two feature vectors of the same size through a fully connected layer. The two feature vectors are fused by splicing to obtain a fused feature Y cat , which is a local deep representation of the harmonic source characteristic time-frequency feature. The splicing operation can be represented as follows:

[0094]

[0095] In the formula: Y cat represents the fused feature after splicing, flatten is a flattening function, and Fc is a fully connected layer function.

[0096] 3.3 BiLSTM network

[0097] The MCNN network can extract local characteristics in a short time period, but it has certain deficiencies in extracting features in a long time period. In order to deeply mine the periodic variation law of harmonic data in a long time scale, the BiLSTM network is used to improve the time sequence performance of the model. The basic unit of the LSTM network is composed of a forgetting gate, an input gate and an output gate, which selectively forgets the information in the cell state and remembers the new information, thereby effectively transmitting useful information for subsequent time calculation and discarding useless information. BiLSTM is composed of two oppositely connected LSTMs, of which the forward LSTM is used to extract information before the sequence data, and the reverse LSTM is responsible for reverse processing of the data. The hidden layer fuses the forward and reverse information, uses the past and future states to enhance the useful information of the network, so as to accurately predict the extracted features. The input of the BiLSTM network is the fused feature Y cat obtained in section 3.2, and the output of the model is the harmonic current X out , which is specifically:

[0098] X out = F BiLSTM (Y cat ) (32)

[0099] In the formula: F BiLSTM represents the BiLSTM network.

[0100] 4 Harmonic source modeling method flow

[0101] (1) The harmonic source voltage and current data collected by the synchronous phasor measurement device are normalized and subjected to FFT analysis. The harmonic component content rate is calculated, and the harmonic principal component frequency of the harmonic source is determined by the threshold judgment rule specified in formula (4).

[0102] (2) Extract the harmonic voltage amplitude and phase angle corresponding to the harmonic principal component frequency of the harmonic source to obtain a harmonic source frequency domain feature set. The frequency domain feature set is reconstructed by IDFT to form a harmonic source frequency domain feature component matrix, and the original voltage sequence collected by the synchronous phasor measurement device is spliced to obtain a time-frequency feature component matrix. The component matrix is converted into a grayscale image by using the matrix normalization method, and is converted into a three-channel time-frequency feature component color image by using the pseudo-color coding technology.

[0103] (3) As shown in Figure 1 , the harmonic source voltage data collected by the synchronous phasor measurement device and the time-frequency feature component color image generated in the above steps are used as the input of the MCNN-BiLSTM model, and the corresponding harmonic source current data collected by the synchronous phasor measurement device is used as the output of the MCNN-BiLSTM model. The network model is continuously trained by a large amount of measured data to realize accurate mapping of the relationship between the harmonic voltage and the harmonic current of the harmonic source, and a harmonic source data-driven model is obtained.

[0104] Those skilled in the art will appreciate that embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0105] The present application is described with reference to flowcharts and / or block diagrams according to the methods, devices (systems), and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the computer or other programmable data processing apparatus produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks

[0106] These computer program instructions can also be stored in a computer-readable memory that can cause the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction devices that implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks Figure 1 The functions specified in one or more flows and / or blocks

[0107] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable devices to generate computer-implemented processes, so that the instructions executed on the computer or other programmable devices provide a process for implementing the flowchart Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0108] It should be noted that, unless otherwise defined, technical terms or scientific terms used in the present application should be understood as their common meanings to those skilled in the art to which the present application pertains. The terms "first", "second" and similar terms used in the present application do not denote any order, quantity or importance, but are only used to distinguish different components. The terms "include" or "contain" and similar terms mean that the elements or objects before the terms encompass the elements or objects listed after the terms and their equivalents, without excluding other elements or objects. The terms "connect" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "upper", "lower", "left", "right" and the like are only used to represent relative positional relationships, which can change accordingly when the absolute positions of the described objects change.

[0109] It should be noted that, unless otherwise defined, technical terms or scientific terms used in the present application should be understood as their common meanings to those skilled in the art to which the present application pertains. The terms "first", "second" and similar terms used in the present application do not denote any order, quantity or importance, but are only used to distinguish different components. The terms "include" or "contain" and similar terms mean that the elements or objects before the terms encompass the elements or objects listed after the terms and their equivalents, without excluding other elements or objects. The terms "connect" or "connected" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms "upper", "lower", "left", "right" and the like are only used to represent relative positional relationships, which can change accordingly when the absolute positions of the described objects change.

[0110] The present patent is not limited to the above best mode, and anyone can derive other various forms of a harmonic source modeling method based on time-frequency feature extraction under the inspiration of the present patent. Any equivalent changes and modifications made within the scope of the patent application of the present patent shall be covered by the present patent.

Claims

1. A harmonic source modeling method based on time-frequency feature extraction, characterized in that: First, the harmonic source voltage and current data acquired by the synchronous phasor measurement device are normalized and analyzed by FFT; the content rate of each harmonic component is calculated, and the frequency of the principal harmonic component of the harmonic source is determined by the set threshold judgment rule. Then, the harmonic voltage amplitude and phase angle corresponding to the frequency of the principal harmonic components of the harmonic source are extracted to obtain the frequency domain feature set of the harmonic source; The frequency domain feature set is reconstructed by IDFT to form the harmonic source frequency domain feature component matrix, and then spliced ​​with the original voltage sequence acquired by the synchronous phasor measurement device to obtain the time-frequency feature component matrix. The component matrix is ​​converted into a grayscale image using matrix normalization, and then converted into a three-channel time-frequency feature component color image using pseudo-color coding technology. Finally, the harmonic source voltage data and the time-frequency characteristic component color map acquired by the synchronous phasor measurement device are used as inputs to the MCNN-BiLSTM model, and the corresponding harmonic source current data acquired by the synchronous phasor measurement device are used as outputs to the MCNN-BiLSTM model. The network model is continuously trained through measured data until the accurate mapping of the relationship between harmonic source harmonic voltage and harmonic current is achieved, thus obtaining the harmonic source data-driven model. The MCNN-BiLSTM model is a harmonic source voltage-current mapping model composed of a combination of multiple convolutional neural network (MCNN) and bidirectional long short-term memory network (BiLSTM). In the MCNN-BiLSTM model, the MCNN model extracts features jointly by a time-domain feature extraction module and a frequency-domain feature extraction module. The two-way parallel structure enables the simultaneous extraction of different feature information from multiple channels and multiple dimensions of input. The MCNN and BiLSTM are connected in series, and the time-frequency domain fusion features extracted by the MCNN model are input into the BiLSTM network to obtain the feature information of harmonic voltage in the entire time domain. The specific steps for harmonic principal component analysis of harmonic sources are as follows: First, the raw data acquired by the synchronous phasor measurement device is normalized to unify the scale of the raw measurement data of the harmonic source: In the formula: The normalized voltage and current sequence is given, and U and I are the original voltage and current sequences measured by the synchronous phasor measurement device. max I max These are the extreme values ​​of the original voltage and current sequence measured by the synchronous phasor measurement device; Secondly, a fast Fourier transform is performed on the normalized voltage and current sequence to obtain the components of the corresponding k-th harmonic: In the formula: The voltage amplitude and phase angle of the kth harmonic are given. The amplitude and phase angle of the k-th harmonic current; Then, the harmonic voltage content HRU is calculated separately. k Harmonic current content (HRI) k : In the formula: U1 and I1 are the amplitudes of the fundamental voltage and fundamental current, respectively; Finally, select all HRUs according to the following formula. k and HRI k Harmonic frequencies that each account for more than 1% are taken as principal harmonic component frequencies, forming the set of principal harmonic component frequencies γ of the harmonic source.

2. The harmonic source modeling method based on time-frequency feature extraction according to claim 1, characterized in that: Based on the set of principal harmonic components, the voltage amplitude and phase angle of the corresponding kth harmonic are extracted to obtain the frequency domain feature set of the harmonic source. The steps to convert F into a time-frequency characteristic component color map are as follows: First, the inverse Fourier transform is used to reconstruct each harmonic voltage component: In the formula: U k (t) represents the characteristic component of the kth harmonic voltage, and N is the number of sampling points; Secondly, after inverse Fourier transform, the harmonic voltage characteristic components {U1(t),U3(t),…,U...} corresponding to the principal harmonic components are obtained. k The time-frequency characteristic component matrix {U1(t), U3(t), ..., U2(t)} is concatenated with the original voltage sequence U(t) acquired by the synchronous phasor measurement device to obtain the time-frequency characteristic component matrix {U1(t), U3(t), ..., U2(t)}. k (t),U(t)}; Then, the time-frequency feature component matrix is ​​processed using the following formula to normalize its data to the range of (0,255), and then it is mapped to grayscale values ​​to obtain the grayscale image of the time-frequency features; In the formula: G(m,n) represents the value of the grayscale matrix in the m-th row and n-th column; x i x represents the value of the i-th data point; min With x max These represent the minimum and maximum values ​​in the time-frequency characteristic component matrix, respectively. Finally, pseudo-color coding is used to convert the grayscale image into a three-channel RGB color image, resulting in a time-frequency feature component color map.

3. The harmonic source modeling method based on time-frequency feature extraction according to claim 1, characterized in that: The temporal feature extraction module adopts a two-layer 1D convolutional network structure. Each 1D convolutional network layer contains a 1D convolutional layer, a 1D regularization layer, and a 1D pooling layer. The harmonic source voltage data X1 acquired by the synchronous phasor measurement device is used as input and sequentially passed through the two 1D convolutional networks to finally obtain the feature vector Y. 1DCNN : Y 1DCNN =F 1DCNN (X1) (7) In the formula: F 1DCNN This indicates the temporal feature extraction module.

4. The harmonic source modeling method based on time-frequency feature extraction according to claim 1, characterized in that: The frequency domain feature learning module employs a two-layer 2D convolutional network. Each layer of the 2D convolutional network contains a 2D convolutional layer, a 2D regularization layer, and a 2D pooling layer. The time-frequency feature component color image X2 is taken as input and sequentially passed through the two 2D convolutional networks to finally obtain the feature vector Y. 2DCNN : Y 2DCNN =F 2DCNN (X2) (8) In the formula: F 2DCNN This indicates the frequency domain feature extraction module.

5. The harmonic source modeling method based on time-frequency feature extraction according to claim 1, characterized in that: The feature vectors output by the time-domain feature extraction module and the frequency-domain feature extraction module are flattened and then converted into two feature vectors of the same size through a fully connected layer. The two feature vectors are then fused by concatenation to obtain the fused feature Y. cat As a local deep representation of the time-frequency characteristics of harmonic sources, the splicing operation is represented as follows: In the formula: Y cat This represents the fused features after splicing, where flatten is the flattening function and Fc is the fully connected layer function.

6. The harmonic source modeling method based on time-frequency feature extraction according to claim 1, characterized in that: The input to the BiLSTM network is the fused feature Y. cat The model output is the harmonic current X. out Specifically: X out =F BiLSTM (Y cat ) (11) In the formula: F BiLSTM This indicates a BiLSTM network.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The computer program includes a harmonic source data-driven model obtained by a harmonic source modeling method based on time-frequency feature extraction as described in any one of claims 1-6.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program includes a harmonic source data-driven model obtained by a harmonic source modeling method based on time-frequency feature extraction as described in any one of claims 1-6.

Images