Classification and identification method of harmonic source load for industrial users

Abstract

The present invention proposes a method for classifying and identifying harmonic source loads of industrial users, which first constructs a multi-harmonic source negative parallel impedance model; secondly, estimates the number of source signals; and finally classifies and identifies harmonic source loads of industrial users based on blind source separation of integrated impedance signals. The signal number estimation result obtained by the method can provide a basis for determining the number of channels for blind source separation of integrated impedance signals. A typical harmonic source load impedance signal frequency feature library is constructed, and the frequency features of different types of harmonic source load impedance signals contained in the feature library can effectively distinguish the types of harmonic source loads. By comparing the results of blind source separation of integrated impedance signals with the frequency feature library of typical harmonic source load impedance signals, the types of harmonic source loads contained in industrial users can be quickly and accurately identified.

Description

Technical Field

[0001] The invention belongs to the technical field of harmonic source identification, and in particular relates to a method for classifying and identifying harmonic source loads of industrial users. Background Art

[0002] With the transformation of energy structure and the rapid improvement of national economic industrialization, the number of common industrial harmonic source loads such as arc furnaces, large-scale converters, and induction heating equipment continues to increase, and the resulting harmonic pollution is becoming increasingly serious.

[0003] The nonlinear equipment of industrial users is the main harmonic source load in the power grid. Detecting and identifying it will not only help improve the load control level of power grid companies, but also provide important technical support for analyzing users' electricity service needs and formulating power quality management plans.

[0004] Typical harmonic source loads for industrial users are mainly divided into two categories: arc type and power electronic type. Arc type loads generally include arc furnaces, electric welders, etc. The harmonics of this type of load come from its internal arc working characteristics; power electronic equipment generally includes induction heating furnaces, frequency converters, etc. The power quality problems caused by this type of equipment to the power grid are mainly affected by the rectifier part. Different types of harmonic sources have different corresponding harmonic control measures. Harmonic source classification and identification is an important basis for formulating targeted harmonic control plans, improving control effects and reducing costs.

[0005] The current harmonic source identification methods can be generally divided into two categories: one is based on the equivalent circuit model, and the other is based on the harmonic state estimation method. The former establishes an equivalent model to divide the system side and the user side at the common connection point, and determines the main harmonic source side based on different positioning theories. The latter establishes a parameter matrix based on the network topology and component harmonic parameters, solves the harmonic state estimation equation, and identifies the harmonic source based on the positive and negative active power injected by the harmonic. Although the above methods can determine whether the user has a harmonic source load, they cannot distinguish the specific type of harmonic source load contained in the user. Summary of the invention

[0006] In view of the defects and shortcomings of the prior art, the present invention mainly proposes two design points to solve the problems:

[0007] (1) Based on the linear superposition relationship between the equivalent impedance of the user bus comprehensive load and the equivalent impedance of each branch load, a method for estimating the number of source impedance signals that constitute the comprehensive equivalent impedance signal is proposed.

[0008] (2) Based on the above signal number estimation results, a method for classifying and identifying harmonic source loads of industrial users based on blind source separation of comprehensive impedance signals is proposed.

[0009] Based on the linear superposition relationship between the equivalent impedance of the user bus comprehensive load and the equivalent impedance of each branch load, a method for estimating the number of source impedance signals that constitute the comprehensive equivalent impedance signal is proposed. The signal number estimation result obtained by this method can provide a basis for determining the number of blind source separation channels of the comprehensive impedance signal. A typical harmonic source load impedance signal frequency feature library is constructed. The frequency features of different types of harmonic source load impedance signals contained in this feature library can effectively distinguish the types of harmonic source loads. By comparing the blind source separation results of the comprehensive impedance signal with the typical harmonic source load impedance signal frequency feature library, the harmonic source load types contained in industrial users can be quickly and accurately identified.

[0010] The technical solution adopted by the present invention to solve the technical problem is:

[0011] A method for classifying and identifying harmonic source loads of industrial users, characterized by:

[0012] Firstly, a negative parallel impedance model of multiple harmonic sources is constructed; secondly, the number of source signals is estimated; finally, the harmonic source loads of industrial users are classified and identified based on blind source separation of comprehensive impedance signals.

[0013] Furthermore, the construction of the multi-harmonic source negative parallel impedance model is specifically as follows:

[0014] (1) Harmonic source load equivalent time domain circuit model

[0015] The load model is represented by the equivalent impedance of resistance and inductance in series. Let R and L be the parameters of the resistive and inductive components in the load equivalent impedance parameters respectively; i(t) and u(t) are the instantaneous values ​​of current and voltage measured at the monitoring point at the user's incoming line respectively; E s and Z s are the equivalent voltage source and equivalent impedance on the system side respectively;

[0016] The relationship between the voltage and current at the load port is:

[0017]

[0018] When the sampling frequency is much greater than the change frequency of the load parameters, it is considered that R and L at two consecutive adjacent sampling points t1 and t2 within a micro-element time Δt are constant. After sorting, we get:

[0019]

[0020] The voltage and current in formula (2) are obtained by the monitoring device, and their derivatives are obtained by calculation; on this basis, the singular value decomposition is used to solve formula (2) to obtain the impedance parameters;

[0021] (2) Multi-harmonic source load parallel impedance model

[0022] The system is equivalent to a multi-impedance parallel network, assuming R k and X k are the equivalent resistance and equivalent reactance parameters of the kth load, k = 1, 2, ..., n; i k is the current of the kth load branch; n is the number of branches;

[0023] According to the circuit characteristics of the parallel network, the relationship between voltage, current and impedance is established as follows:

[0024]

[0025] Where: They are the voltage and current integrated vectors of the monitoring point respectively; is the comprehensive current vector of the kth branch;

[0026] make According to the impedance parallel calculation formula, we can get from formula (3):

[0027]

[0028] In formula (4) is the proportionality coefficient between the impedance of the kth branch and the total impedance of the load, and the sum of the reciprocals of the proportionality coefficients is 1.

[0029] Furthermore, the estimating the number of source signals is specifically as follows:

[0030] First, the comprehensive impedance signals R(t) and X(t) of the multi-harmonic source negative parallel impedance model are decomposed by EEMD; the signal intrinsic mode function containing the characteristics of the source signal is obtained; and the IMF with similar average frequency is then i The reconstructed virtual observation signal is obtained by superposition, and the reconstructed virtual observation signal and the original signal are further processed by removing the mean to form a new multi-channel observation signal, so that the number of observation signals is greater than the number of source signals;

[0031] Secondly, the singular value decomposition of the autocorrelation matrix of the multi-channel observation signal y(t) is performed;

[0032]

[0033] Where: (.) H represents complex conjugate transformation; V s is the principal component arranged in descending order after the signal is decomposed; V N Corresponding to the decomposed noise; Λ S =diag{λ1≥λ2…≥λ s} and Λ N =diag{λ s+1 ,λ s+2 ,…,λh} is the main component of the signal V s and the noise signal V N The corresponding eigenvalue of ;

[0034] The interface between the source signal and the noise is determined by comparing the ratio of adjacent eigenvalues, and the index corresponding to the maximum ratio δ is taken as the interface between the source signal and the noise;

[0035] δ=max(σ1 / σ2,σ2 / σ3,…,σ k-1 / σ k ) (9)

[0036] Define the source signal and noise interface index plus 1 as the number of actual source impedance signals W:

[0037] W = δ + 1 (10).

[0038] Furthermore, the classification and identification of harmonic source loads of industrial users based on blind source separation of comprehensive impedance signals is specifically as follows:

[0039] 1. Virtual channel signal construction

[0040] The integrated impedance signals R(t), X(t) are decomposed by EEMD, and the correlation coefficients of the IMF components in the intrinsic mode are calculated. The IMF components with correlation coefficients greater than 0.8 are set as strong correlation components to be directly used as one virtual channel signal. The IMF components with correlation coefficients less than 0.8 are combined with the strong correlation components as virtual channel signals to ensure that the constructed virtual channel signal can contain more characteristic information of the source signal. The correlation coefficient γ is obtained by formula (11):

[0041]

[0042] Where: cov() represents the covariance function; var() represents the variance function. The larger the γ value, the better it reflects the mutual matching degree of the two signals.

[0043] After calculating the correlation coefficient γ, the virtual channel signal and the original signal are recombined into a multi-channel observation signal z(t). The number of channels is estimated based on the source impedance signal, that is:

[0044] z(t)=[x(t),z1(t),z2(t),…,z δ (t)] (12)

[0045] Where: z i (t) represents the i-th virtual channel signal, i=1, 2, ..., δ;

[0046] 2. Classification and identification of harmonic source loads

[0047] The FastICA algorithm is used to perform blind source separation on the recombined signal to obtain the reactance signal of each branch Y(t)=[Y1(t),Y2(t),...,Y W (t)], and further perform spectrum analysis on Y(t), match it with the typical harmonic source load impedance signal frequency feature library to identify the harmonic source load type.

[0048] Compared with the prior art, the present invention and its preferred embodiment propose a method for estimating the number of source impedance signals constituting a comprehensive equivalent impedance signal based on the linear superposition relationship between the user bus comprehensive load equivalent impedance and the branch load equivalent impedance. The signal number estimation result obtained by this method can provide a basis for determining the number of blind source separation channels for the comprehensive impedance signal. A typical harmonic source load impedance signal frequency feature library is constructed. The frequency features of different types of harmonic source load impedance signals contained in the feature library can effectively distinguish the types of harmonic source loads. By comparing the blind source separation results of the comprehensive impedance signal with the typical harmonic source load impedance signal frequency feature library, the harmonic source load types contained in industrial users can be quickly and accurately identified. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0050] Figure 1 Schematic diagram of load equivalent time domain circuit model according to an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of a multi-load equivalent circuit model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0053] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0054] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0055] The following is a further detailed description of this embodiment with reference to the accompanying drawings:

[0056] The following is a more specific introduction to the present invention in conjunction with the accompanying drawings using an embodiment: 1 Harmonic source load equivalent impedance model

[0057] The essential reason for harmonic generation by harmonic sources is that their load equivalent impedance changes nonlinearly with the amount of electricity. The multi-harmonic source negative parallel impedance model is the basis for classification and identification of harmonic source loads of industrial users.

[0058] (1) Harmonic source load equivalent time domain circuit model

[0059] Since most industrial loads are inductive, the load model can be represented by the equivalent impedance of resistance and inductance in series, such as Figure 1 In the figure: R and L are the parameters of the resistive and inductive components in the load equivalent impedance parameters; i(t) and u(t) are the instantaneous values ​​of current and voltage measured at the monitoring point at the user's incoming line; E s and Z s are the equivalent voltage source and equivalent impedance on the system side respectively.

[0060] Depend on Figure 1 It can be seen that the relationship between the voltage and current at the load port is:

[0061]

[0062] When the sampling frequency is much greater than the change frequency of the load parameters, it can be considered that R and L at two consecutive adjacent sampling points t1 and t2 within a micro-element time Δt are unchanged. After sorting, we can get:

[0063]

[0064] The voltage and current in formula (2) can be obtained by the monitoring device, and their derivatives can be derived from the monitoring data through relevant algorithms. On this basis, the SVD method can be used to solve formula (2) to obtain the impedance parameters.

[0065] (2) Multi-harmonic source load parallel impedance model

[0066] Most loads in the power system are inductive and connected in parallel. Therefore, the system can be equivalent to a multi-impedance parallel network, such as Figure 2 As shown in the figure: R k and L k are the equivalent resistance and equivalent inductance parameters of the kth load, k = 1, 2, ..., n; i k is the current of the kth load branch; n is the number of branches.

[0067] like Figure 2 As shown, the relationship between voltage, current and impedance is established based on the circuit characteristics of the parallel network:

[0068]

[0069] Where: They are the voltage and current integrated vectors of the monitoring point respectively; is the comprehensive current vector of the kth branch.

[0070] make According to the impedance parallel calculation formula, it can be obtained from formula (3):

[0071]

[0072] In formula (4) is the proportionality coefficient between the impedance of the kth branch and the total impedance of the load, and the sum of the reciprocals of the proportionality coefficients is 1.

[0073] 2 Source signal number estimation method

[0074] Perform EEMD decomposition on the above-mentioned integrated impedance signals R(t), X(t) (taking X(t) as an example) to obtain the signal intrinsic mode function X containing the characteristics of the source signal IMF =[IMF1,IMF2,...,IMF n ,R n ]. Since the decomposed intrinsic mode function IMF i contains the characteristic scale information and characteristic frequency information of the signal. In this embodiment, IMFs with similar average frequencies are i The reconstructed virtual observation signal X is obtained by superposition IMF ′=[IMF′1,IMF′2,...,IMF′ n ,R′ n ], further X IMF ′ and the original signal X(t) are processed by removing the mean to form a new multi-channel observation signal Make the number of observed signals greater than the number of source signals.

[0075] Secondly, the autocorrelation matrix of the multi-channel observation signal y(t) is subjected to singular value decomposition.

[0076]

[0077] Where: (.) H Denotes the complex conjugate transformation. V s is the principal component arranged in descending order after the signal is decomposed; V N Corresponding to the decomposed noise; Λ S =diag{λ1≥λ2…≥λ s} and Λ N =diag{λ s+1 ,λ s+2 ,…,λh} is the main component of the signal V s and the noise signal V N The corresponding eigenvalue of .

[0078] The eigenvalue of the source signal subspace will be significantly greater than the eigenvalue of the noise subspace. The method of this embodiment determines the interface between the source signal and the noise by comparing the ratio of adjacent eigenvalues, and takes the index corresponding to the maximum ratio δ as the interface between the source signal and the noise.

[0079] δ=max(σ1 / σ2,σ2 / σ3,…,σ k-1 / σ k ) (9)

[0080] Considering that in the actual industrial user load equivalent impedance model, the impedance signal of the linear load will have a small degree of nonlinear fluctuation due to the influence of background harmonics, and its variance after de-averaging will be significantly smaller than the harmonic source load impedance signal, approximating the noise characteristics, the method of this embodiment defines the source signal and noise interface index plus 1 as the number of actual source impedance signals W:

[0081] W = δ + 1 (10).

[0082] 3 Classification and identification method of industrial user harmonic source load based on blind source separation of comprehensive impedance signal

[0083] 3.1 Virtual channel signal construction

[0084] Perform EEMD decomposition on the integrated impedance signal R(t), X(t) (taking X(t) as an example) and calculate the intrinsic mode function X IMF =[IMF1,IMF2,...,IMF n ,R n ] The correlation coefficient of each IMF component in the embodiment is set as the strong correlation component. The IMF component with a correlation coefficient greater than 0.8 can be directly used as a virtual channel signal, and the IMF component with a correlation coefficient less than 0.8 should be combined with the strong correlation component as a virtual channel signal to ensure that the constructed virtual channel signal can contain more of all the characteristic information of the source signal. The correlation coefficient γ is obtained by formula (11):

[0085]

[0086] Where: cov() represents the covariance function; var() represents the variance function. The larger the γ value, the more it can reflect the degree of mutual matching between the two signals.

[0087] After calculating the correlation coefficient γ, the virtual channel signal and the original signal are recombined into a multi-channel observation signal z(t) according to the above rules. The number of channels is estimated based on the above-mentioned section source impedance signal, that is:

[0088] z(t)=[x(t),z1(t),z2(t),…,z δ (t)] (12)

[0089] Where: z i (t) represents the i-th virtual channel signal, i=1, 2, ..., δ.

[0090] 3.2 Harmonic source load classification and identification method

[0091] The FastICA algorithm is used to perform blind source separation on the recombined signal to obtain the reactance signal of each branch Y(t)=[Y1(t),Y2(t),...,Y W (t)], further perform spectrum analysis on Y(t), match it with the typical harmonic source load impedance signal frequency feature library to identify the harmonic source load type. Typical harmonic source loads of industrial user loads are mainly divided into two types: arc type and power electronic type. Arc type loads generally include arc furnaces, electric welding machines, etc., and power electronic equipment generally includes induction heating furnaces, inverters, etc. The method of this embodiment constructs a load impedance signal frequency feature library for arc furnaces and different types of rectifiers.

[0092] (1) Rectifier

[0093] The rectifier is the main source of harmonics generated by power electronic loads, and the on and off processes of the power electronic devices inside the rectifier will inject harmonics into the power grid. The frequency characteristics of the rectifier impedance signal proposed by the method of this embodiment are shown in Table 1.

[0094]

[0095] (2) Electric Arc Furnace

[0096] The main harmonics of arc-type loads represented by arc furnaces come from arcs generated during load operation. When the arc length and arc burning are in a relatively stable state, arc-type loads will inject periodic harmonics into the power grid. The method in this embodiment defines the low-frequency component in the (0, 50Hz) interval in the impedance signal as the impedance signal feature of the arc furnace.

[0097] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0098] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of industrial user harmonic source load classification and identification methods under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. A method for classifying and identifying harmonic source loads of industrial users, characterized by: S1. Construct a multi-harmonic source negative parallel impedance model; S2, estimating the number of source signals; S3, classify and identify harmonic source loads of industrial users based on blind source separation of comprehensive impedance signals; The constructing of the multi-harmonic source negative parallel impedance model comprises: S1.1、Time domain circuit model of harmonic source and load equivalent; The load model is represented by the equivalent impedance of resistance and inductance in series. Let R and L represent the equivalent resistance and reactance of the load respectively; i(t) and u(t) are the instantaneous values ​​of current and voltage measured at the monitoring point at the user's incoming line respectively; E s and Z s are the equivalent voltage source and equivalent impedance on the system side respectively; The relationship between the voltage and current at the load port is: (1) When the sampling frequency is much greater than the frequency of change of the load parameters, it is considered that R and L at two consecutive adjacent sampling points t1 and t2 within a micro-element time Δt are constant. After sorting, we get: (2) The voltage and current in formula (2) are obtained by the monitoring device, and their derivatives are obtained by calculation; The impedance parameters are obtained by solving equation (2) using singular value decomposition; S1.2, multi-harmonic source load parallel impedance model; The system is equivalent to a multi-impedance parallel network, assuming R k and X k are the equivalent resistance and equivalent reactance parameters of the kth load, k = 1, 2, ..., n; i k is the current of the kth load branch; n is the number of branches; According to the circuit characteristics of the parallel network, the relationship between voltage, current and impedance is established as follows: (3) Where: , They are the voltage and current integrated vectors of the monitoring point respectively; is the comprehensive current vector of the kth branch; make According to the impedance parallel calculation formula, we can get from formula (3): (4) In formula (4), is the proportionality coefficient between the impedance of the kth branch and the total impedance of the load, and the sum of the reciprocals of the proportionality coefficients is 1; The estimating the number of source signals comprises: S2.

1. Comprehensive impedance signal R of negative parallel impedance model of multi-harmonic source k (t), X k (t) Perform EEMD decomposition; obtain the signal intrinsic mode function containing the characteristics of the source signal; then use the IMF with similar average frequency i The reconstructed virtual observation signal is obtained by superposition, and the reconstructed virtual observation signal and the original signal are further processed by removing the mean to form a new multi-channel observation signal, so that the number of observation signals is greater than the number of source signals; S2.

2. Perform singular value decomposition on the autocorrelation matrix of the multi-channel observation signal y(t): (8) Where: (.) H represents complex conjugate transformation; V s is the principal component arranged in descending order after the signal is decomposed; V N Corresponding to the decomposed noise; Λ S =diag{λ1≥λ2…≥λ s } and Λ N =diag{λ s+1 ,λ s+2 ,…,λ h } is the main component of the signal V s and the noise signal V N The corresponding eigenvalue of ; The interface between the source signal and the noise is determined by comparing the ratio of adjacent eigenvalues, and the index corresponding to the maximum ratio δ is taken as the interface between the source signal and the noise; δ=max(σ1 / σ2,σ2 / σ3,…,σ k-1 / s k ) (9) Define the source signal and noise interface index plus 1 as the number of actual source impedance signals W: Wδ=δ+1 (10); The classification and identification of industrial user harmonic source loads based on comprehensive impedance signal blind source separation is specifically as follows: S3.1, virtual channel signal construction; Perform EEMD decomposition on the integrated impedance signals R(t), X(t) and calculate the IMF in the eigenmode i The size of the correlation coefficient of the components; set the IMF with a correlation coefficient greater than 0.8 i The component is a strongly correlated component, which can be directly used as a virtual channel signal, and the IMF with a correlation coefficient less than 0.8 i The component is combined with the strongly correlated component as a virtual channel signal to ensure that the constructed virtual channel signal can contain more characteristic information of the source signal; the correlation coefficient γ is obtained by formula (11): （11） Where: cov() represents the covariance function; var() represents the variance function. The larger the γ value, the better it reflects the mutual matching degree of the two signals. After calculating the correlation coefficient γ, the virtual channel signal and the original signal are recombined into a multi-channel observation signal z(t). The number of channels is estimated based on the source impedance signal, that is: z(t)=[x(t),z1(t),z2(t),…,z δ (t)] (12) Where: z i (t) represents the i-th virtual channel signal, i=1, 2, ..., δ; S3.2, Classification and identification of harmonic source loads; The FastICA algorithm is used to perform blind source separation on the recombined signal to obtain the reactance signal of each branch Y(t)=[Y1(t),Y2(t),...,Y W (t)], and further perform spectrum analysis on Y(t), match it with the typical harmonic source load impedance signal frequency feature library to identify the harmonic source load type.

Images