High-proportion distributed power supply distribution network short circuit feature extraction method

Abstract

The invention relates to the technical field of power distribution network fault diagnosis, and discloses a high-proportion distributed power supply power distribution network short circuit feature extraction method, which comprises the following steps: S1, constructing a successive multivariate variational mode decomposition algorithm; s2, performing parameter optimization on the successive multivariate variational mode decomposition algorithm by adopting a weighted chimpanzee optimization algorithm, and determining a maximum decomposition layer number and a maximum penalty factor required by signal decomposition by utilizing the successive multivariate variational mode decomposition algorithm; s3, roughly positioning the fault area; s4, using MOMEDA to extract short circuit current fault features; according to the method, the weighted chimpanzee optimization algorithm is adopted, the successive multivariate variational mode decomposition algorithm is subjected to parameter optimization, the maximum layer number and the maximum penalty factor required for decomposing the short-circuit signal are obtained, and the method can be used for decomposing multichannel signals, so that the accuracy of decomposing the short-circuit current signal is improved, and excessive fault signal components are prevented from being screened out.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network fault diagnosis, and in particular to a method for extracting short-circuit features of a high-proportion distributed power distribution network. Background Art

[0002] The vigorous development of new energy power generation can reduce the primary energy consumed during the operation of the power system, providing more electricity for the power system while achieving environmental protection.

[0003] New energy is a renewable energy source, and their common characteristics are: low energy density, dispersed, intermittent, and random. At present, the access of new energy is mainly divided into three categories: off-grid storage, grid-connected power generation, and a hybrid system of the two. New energy grid-connected power generation has certain requirements on the acceptance capacity of the distribution network. Therefore, the grid connection of new energy power sources may cause a series of negative impacts on the power system.

[0004] New energy sources are generally connected to the power grid through converters and a large number of power electronic devices, and their output is affected by the weather, with significant intermittent and fluctuating characteristics, which will cause certain DC components and harmonic components in the system, thereby causing grid voltage distortion and affecting the power quality of the system. At the same time, the presence of these harmonic components may also cause large measurement errors in the measuring instruments in the system, threatening the correct operation of relay protection and automatic devices.

[0005] New energy sources have characteristics such as state diversity, distributed layout, and output nonlinearity, which make it difficult to predict and determine the short-circuit current when a distribution network fails. The analysis of short-circuit characteristics becomes more complicated, affecting the reliability of short-circuit fault diagnosis and removal.

[0006] Considering that complex fault signals often contain a lot of redundancy and interference, how to accurately extract effective fault features is the key to efficiently distinguish the type of fault. Traditional signal feature extraction uses Fourier transform, wavelet transform, empirical mode decomposition and other methods. These methods have certain limitations when processing nonlinear noisy fault signals. In order to obtain more accurate fault features, the SVMMD algorithm can be combined with the intelligent optimization algorithm, and the artificial intelligence optimization algorithm can be used to optimize the algorithm parameters for signal decomposition, which can effectively improve the accuracy of signal decomposition and extraction.

[0007] Based on this, in order to accurately extract the fault feature information contained in the short-circuit current signal of a high-proportion distributed power distribution network, this paper proposes a short-circuit feature extraction method for a high-proportion distributed power distribution network based on the weighted chimpanzee-SMVMD optimization algorithm. The short-circuit current signal is decomposed by SVMMD and the short-circuit fault features are extracted based on MOMEDA to improve the accuracy of fault information extraction. Summary of the invention

[0008] The purpose of the present invention is to provide a method for extracting short-circuit features of a high-proportion distributed power distribution network to solve the problems raised in the above-mentioned background technology.

[0009] In order to achieve the above object, the present invention provides the following technical solution: a method for extracting short-circuit characteristics of a high-proportion distributed power distribution network, comprising the following steps:

[0010] S1. Construct a successive multivariate variational mode decomposition algorithm;

[0011] S2. Using the weighted chimpanzee optimization algorithm to optimize the parameters of the successive multivariate mode decomposition algorithm, determine the maximum number of decomposition layers and the maximum penalty factor required to obtain signal decomposition using the successive multivariate mode decomposition algorithm;

[0012] S201, setting the search range of the maximum decomposition level k and the maximum penalty factor α, k = [2, 15], α = [1000, 5000];

[0013] S202, initialize the weighted chimpanzee optimization algorithm, use k and α as the coordinates of the chimpanzee's location, randomly generate 10 chimpanzees within the search range, and take the maximum number of iterations T=20;

[0014] S203, the average E of the components decomposed by the multivariate mode decomposition algorithm of each channel c The inverse of the sum is used as the fitness function. The smaller this value is, the better the chimpanzee's position is.

[0015] S204, using the chasing, attacking, and searching habits of the chimpanzee population, updating the new population until the number of iterations is completed and the best chimpanzee position is obtained, which is the best combination of the maximum decomposition layer number k and the maximum penalty factor α;

[0016] S3, online identification of short-circuit currents in different areas, using current integral equivalent to reflect the severity of the fault, roughly locating the fault area, and further extracting the characteristics of the short-circuit current signal in the area;

[0017] S4. Decompose the current signal in the rough fault area by using the successive multivariate variational mode decomposition algorithm, and extract the short-circuit current fault characteristics by using MOMEDA for the decomposed current signal.

[0018] Before constructing the successive multivariate variational mode decomposition algorithm, a multi-channel signal x(t) is first constructed:

[0019]

[0020] Where: is the kth mixing vector, u k (t) is the separated kth IMF component; n(t) is white Gaussian noise;

[0021] In order to continuously extract the current component, the multi-channel signal is decomposed into two parts, namely:

[0022]

[0023] Where: x r (t) is the residual signal.

[0024] Among them, on the basis of constructing multi-channel signals, the constraints required for the successive multivariate variational mode decomposition algorithm are further clarified and established:

[0025]

[0026] Where: α is the penalty factor, λ(t), γ k is the Lagrange multiplier, <·> is the inner product calculation, x(t) is the unprocessed part of the signal, and J1, J2, and J3 are spectrum overlap constraints.

[0027] Among them, using the multiplier alternating direction method, u k (t), ω(t), λ c (t) Iterate to find the optimal solution, and the model is updated as follows:

[0028]

[0029] Where: is the signal after Fourier transformation, n is the number of iterations;

[0030] The center frequency is updated as follows:

[0031]

[0032] Update accordingly for:

[0033]

[0034]

[0035] Where: α=min{γα,α max};

[0036] Set the iteration cutoff condition to:

[0037] Or k = K

[0038] Among them: ε1, ε2 are set constants.

[0039] The envelope spectrum peak factor is used to describe the ratio between the amplitude of the signal's envelope spectrum at the peak and the average value, E c The larger the value, the better the signal periodicity, so define E c for:

[0040]

[0041] Where: max(X(j)) is the maximum value in the frequency range from 0 to half of the sampling frequency in the envelope spectrum, and M is half of the total number of sampling points.

[0042] Among them, the equivalent charge is obtained by integrating the current, and the current integral expression is:

[0043]

[0044] Where: q i,j represents the equivalent charge passing through line i, j, l i,j represents the length of line i,j;

[0045] The real-time identification method is used to calculate the equivalent charge flowing on each line in any time period, which serves as the basis for roughly locating the fault area.

[0046] Among them, the modular function is used to divide the weighted distribution network into regions in combination with the equivalent charge;

[0047] To divide the distribution network into regions, it is necessary to define the modular function first:

[0048]

[0049] Where: q ij is defined as the edge weight between node i and node j. When using the calculated equivalent charge, m is the sum of the weights of all edges in the network, and k is i is the sum of the weights of the edges connected to node i, c i and c j Respectively represent the number of the region where node i and node j are located. If node i and node j are divided into the same region, then δ(c i ,c j ) takes 1, otherwise takes 0;

[0050] The meaning of modularity Q is represented by q ij That is, the larger the Q is, the larger the current inside the area is, and the greater the possibility of a fault in the area is. According to the size of Q, the fault location can be roughly located, and the feature extraction of the short-circuit current signal in the area can be further performed.

[0051] Among them, in the process of extracting the short-circuit current fault characteristics using MOMEDA, the vibration signal expression in the short-circuit current is:

[0052] x n =h×y n +e

[0053] Where: y n is the short-circuit current signal, h is the impulse response function, and e is the noise;

[0054] The actual current signal y after decomposition by the successive multivariate mode decomposition algorithm n for:

[0055]

[0056] Convert the minimum entropy deconvolution method to solve the maximum of multiple D norms:

[0057]

[0058] Where: t is the constant vector that determines the fault location;

[0059] Normalize the MDN and further solve it:

[0060]

[0061] Find the particular solution f and substitute it into y n Get the original accurate short-circuit current component.

[0062] Among them, after the fault current signal is extracted, multiple modal components containing fault characteristic information are obtained.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. The present invention adopts a weighted chimpanzee optimization algorithm to optimize the parameters of the successive multivariate variational mode decomposition algorithm, and obtains the maximum number of layers and the maximum penalty factor required for decomposing the short-circuit signal, which can be used for the decomposition of multi-channel signals, thereby improving the accuracy of short-circuit current signal decomposition and avoiding screening out too many fault signal components;

[0065] 2. The present invention adopts an online automatic identification transient current signal area division strategy, uses the fluctuation characteristics of the signal to roughly locate the fault location, and further extracts short-circuit characteristics and performs fault diagnosis and isolation removal on this basis, which can reduce computer calculations and improve the efficiency of system response, thereby reducing the time from fault occurrence to removal and improving the reliability of the distribution network;

[0066] 3. The present invention is well applicable to distribution networks with a high or low proportion of new energy power sources, and solves the problem that the short-circuit current is difficult to predict and determine when the distribution network fails due to the state diversity, distribution layout characteristics, output nonlinearity and other characteristics of new energy power sources;

[0067] 4. The present invention is based on the successive multivariate variational mode decomposition algorithm and MOMEDA and combines short-circuit influencing factors to extract key distribution network short-circuit characteristics, effectively improve the accuracy of short-circuit feature extraction of high-proportion distributed power distribution networks, and avoid fault information loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The present invention discloses a flow chart of using the weighted chimpanzee optimization algorithm to perform parameter optimization on the successive multivariate variational mode decomposition algorithm. DETAILED DESCRIPTION

[0069] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0070] See also Figure 1 The present invention proposes a short-circuit feature extraction method for a high-proportion distributed power distribution network, which uses a successive multivariate variational mode decomposition algorithm (SMVMD algorithm) to decompose the short-circuit current signal, and extracts the short-circuit fault feature based on MOMEDA to improve the accuracy of fault information extraction, including the following steps:

[0071] S1. Construct a successive multivariate variational mode decomposition algorithm;

[0072] S2. Using the weighted chimpanzee optimization algorithm to optimize the parameters of the successive multivariate mode decomposition algorithm, determine the maximum number of decomposition layers and the maximum penalty factor required to obtain signal decomposition using the successive multivariate mode decomposition algorithm;

[0073] The adopted successive multivariate mode decomposition algorithm (SMVMD algorithm) needs to select the appropriate maximum decomposition layer k and maximum penalty factor α. The selection of parameters will affect the result of signal decomposition. In order to improve the decomposition accuracy and minimize the loss of fault features, it is necessary to optimize the maximum decomposition layer k and the maximum penalty factor α. This application chooses to use the weighted chimpanzee optimization algorithm for parameter optimization. The specific steps are as follows:

[0074] S201, setting the search range of the maximum decomposition level k and the maximum penalty factor α, k = [2, 15], α = [1000, 5000];

[0075] S202, initialize the weighted chimpanzee optimization algorithm, use k and α as the coordinates of the chimpanzee's location, randomly generate 10 chimpanzees within the search range, and take the maximum number of iterations T=20;

[0076] S203, the average E of the components decomposed by the multivariate mode decomposition algorithm of each channel c The inverse of the sum is used as the fitness function. The smaller this value is, the better the chimpanzee's position is.

[0077] S204, using the chasing, attacking, and searching habits of the chimpanzee population, updating the new population until the number of iterations is completed and the best chimpanzee position is obtained, which is the best combination of the maximum decomposition layer number k and the maximum penalty factor α;

[0078] S3. For the distribution network, the short-circuit current at the fault point after the short circuit has obvious characteristics. The farther away from the fault point, the smaller the current. The short-circuit current in different areas is identified online, and the current integral equivalent is used to reflect the severity of the fault. The fault area is roughly located, and the feature extraction of the short-circuit current signal in the area is further performed;

[0079] S4. The current signal in the rough fault area is decomposed by the successive multivariate variational mode decomposition algorithm. The short-circuit current fault characteristics are extracted by MOMEDA for the decomposed current signal to improve the accuracy of fault information extraction.

[0080] Specifically, before constructing the successive multivariate variational mode decomposition algorithm (SMVMD algorithm), it is necessary to construct a multi-channel signal x(t):

[0081]

[0082] Where: is the kth mixing vector, u k (t) is the separated kth IMF component; n(t) is white Gaussian noise;

[0083] In order to continuously extract the current component, the multi-channel signal is decomposed into two parts, namely:

[0084]

[0085] Where: x r (t) is the residual signal.

[0086] Among them, on the basis of constructing multi-channel signals, the constraints required for the successive multivariate variational mode decomposition algorithm are further clarified and established:

[0087]

[0088] Where: α is the penalty factor, λ(t), γ k is the Lagrange multiplier, <·> is the inner product calculation, x(t) is the unprocessed part of the signal, and J1, J2, and J3 are spectrum overlap constraints.

[0089] Specifically, using the multiplier alternating direction method, u k (t), ω(t), λ c (t) Iterate to find the optimal solution, and the model is updated as follows:

[0090]

[0091] Where: is the signal after Fourier transformation, n is the number of iterations;

[0092] The center frequency is updated as follows:

[0093]

[0094] Update accordingly for:

[0095]

[0096]

[0097] Where: α=min{γα,α max};

[0098] Set the iteration cutoff condition to:

[0099] Or k = K

[0100] Among them: ε1, ε2 are set constants.

[0101] Specifically, the envelope spectrum peak factor is used to describe the ratio between the amplitude of the signal envelope spectrum at the peak and the average value, E c The larger the value, the better the signal periodicity, so define E c for:

[0102]

[0103] Where: max(X(j)) is the maximum value in the frequency range from 0 to half of the sampling frequency in the envelope spectrum, and M is half of the total number of sampling points.

[0104] Specifically, the equivalent charge is obtained by integrating the current, and the current integral expression is:

[0105]

[0106] Where: q i,jrepresents the equivalent charge passing through line i, j, l i,j represents the length of line i,j;

[0107] The real-time identification method is used to calculate the equivalent charge flowing on each line in any time period, which serves as the basis for roughly locating the fault area.

[0108] Specifically, the weighted distribution network is divided into regions using modular functions in combination with the equivalent charge;

[0109] To divide the distribution network into regions, it is necessary to define the modular function first:

[0110]

[0111] Where: q ij is defined as the edge weight between node i and node j. When using the calculated equivalent charge, m is the sum of the weights of all edges in the network, and k is i is the sum of the weights of the edges connected to node i, c i and c j Respectively represent the number of the region where node i and node j are located. If node i and node j are divided into the same region, then δ(c i ,c j ) takes 1, otherwise takes 0;

[0112] Specifically, the meaning of modularity Q is represented by q ij That is, the larger the Q is, the larger the current inside the area is, and the greater the possibility of a fault in the area is. According to the size of Q, the fault location can be roughly located, and the feature extraction of the short-circuit current signal in the area can be further performed.

[0113] Specifically, in order to reduce the actual amount of calculation and shorten the fault diagnosis time, the vibration signal expression in the short-circuit current is as follows when extracting the short-circuit current fault characteristics using MOMEDA for the current signal in the rough fault area located:

[0114] x n =h×y n +e

[0115] Where: y n is the short-circuit current signal, h is the impulse response function, and e is the noise;

[0116] The actual current signal y after decomposition by the successive multivariate mode decomposition algorithm n for:

[0117]

[0118] Convert the minimum entropy deconvolution method to solve the maximum of multiple D norms:

[0119]

[0120] Where: t is the constant vector that determines the fault location;

[0121] Normalize the MDN and further solve it:

[0122]

[0123] Find the particular solution f and substitute it into y n Get the original accurate short-circuit current component.

[0124] Specifically, after the fault current signal is extracted, a plurality of modal components containing fault characteristic information (such as system frequency, fault location, etc.) are obtained.

[0125] In summary, the present invention proposes a short-circuit feature extraction method for a high-proportion distributed power distribution network based on a weighted chimpanzee-SMVMD optimization algorithm. The short-circuit current signal is decomposed by a successive multivariate variational mode decomposition algorithm (SMVMD algorithm), and the short-circuit fault features are extracted based on MOMEDA to improve the accuracy of fault information extraction.

[0126] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting short-circuit features of a high-proportion distributed power distribution network, characterized in that: The following steps are involved: S1. Construct a successive multivariate variational mode decomposition algorithm; S2. Using the weighted chimpanzee optimization algorithm to optimize the parameters of the successive multivariate mode decomposition algorithm, determine the maximum number of decomposition layers and the maximum penalty factor required to obtain signal decomposition using the successive multivariate mode decomposition algorithm; S201, setting the search range of the maximum decomposition level k and the maximum penalty factor α, k = [2, 15], α = [1000, 5000]; S202, initialize the weighted chimpanzee optimization algorithm, use k and α as the coordinates of the chimpanzee's location, randomly generate 10 chimpanzees within the search range, and take the maximum number of iterations T=20; S203, the average E of the components decomposed by the multivariate mode decomposition algorithm of each channel c The inverse of the sum is used as the fitness function. The smaller this value is, the better the chimpanzee's position is. S204, using the chasing, attacking, and searching habits of the chimpanzee population, updating the new population until the number of iterations is completed and the best chimpanzee position is obtained, which is the best combination of the maximum decomposition layer number k and the maximum penalty factor α; S3, online identification of short-circuit currents in different areas, using current integral equivalent to reflect the severity of the fault, roughly locating the fault area, and further extracting the characteristics of the short-circuit current signal in the area; S4. Decompose the current signal in the rough fault area by using the successive multivariate variational mode decomposition algorithm, and extract the short-circuit current fault characteristics by using MOMEDA for the decomposed current signal.

2. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 1 is characterized in that: Before constructing the successive multivariate variational mode decomposition algorithm, first construct the multi-channel signal x(t): Where: is the kth mixing vector, u k (t) is the separated kth IMF component; n(t) is white Gaussian noise; In order to continuously extract the current component, the multi-channel signal is decomposed into two parts, namely: Where: x r (t) is the residual signal.

3. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 2 is characterized in that: On the basis of constructing multi-channel signals, the constraints required for the successive multivariate variational mode decomposition algorithm are further clarified and established: Where: α is the penalty factor, λ(t), γ k is the Lagrange multiplier, <·> is the inner product calculation, x(t) is the unprocessed part of the signal, and J1, J2, and J3 are spectrum overlap constraints.

4. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 3 is characterized in that: Using the method of alternating directions of multipliers, k (t), ω(t), λ c (t) Iterate to find the optimal solution, and the model is updated as follows: Where: is the signal after Fourier transformation, n is the number of iterations; The center frequency is updated as follows: Update accordingly for: In the formula: α=min{γα,α max }; Set the iteration cutoff condition to: Or k = K Among them: ε1, ε2 are set constants.

5. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 4 is characterized in that: The envelope spectrum peak factor is used to describe the ratio between the amplitude at the peak and the average value of the envelope spectrum of the signal, E c The larger the value, the better the signal periodicity, so define E c for: Where: max(X(j)) is the maximum value in the envelope spectrum within the frequency range from 0 to half the sampling frequency, and M is half of the total number of sampling points.

6. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 5, characterized in that: The equivalent charge is obtained by integrating the current. The current integral expression is: Where: q i,j represents the equivalent charge passing through line i, j, l i,j represents the length of line i,j; The real-time identification method is used to calculate the equivalent charge flowing on each line in any time period, which serves as the basis for roughly locating the fault area.

7. A method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 6, characterized in that: Combined with the equivalent charge, modular function is used to divide the weighted distribution network into regions. To divide the distribution network into regions, it is necessary to define the modular function first: Where: q ij is defined as the edge weight between node i and node j. When using the calculated equivalent charge, m is the sum of the weights of all edges in the network, and k is i is the sum of the weights of the edges connected to node i, c i and c j Respectively represent the number of the region where node i and node j are located. If node i and node j are divided into the same region, then δ(c i ,c j ) takes 1, otherwise takes 0; The meaning of modularity Q is represented by q ij That is, the larger the Q is, the larger the current inside the area is, and the greater the possibility of a fault in the area is. According to the size of Q, the fault location can be roughly located, and the feature extraction of the short-circuit current signal in the area can be further performed.

8. The method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 7, characterized in that: In the process of extracting the short-circuit current fault characteristics using MOMEDA, the vibration signal expression in the short-circuit current is: x n =h×y n +e Where: y n is the short-circuit current signal, h is the impulse response function, and e is the noise; The actual current signal y after decomposition by the successive multivariate mode decomposition algorithm n for: Convert the minimum entropy deconvolution method to solve the maximum of multiple D norms: Where: t is a constant vector that determines the location of the fault; Normalize the MDN and further solve it: Find the particular solution f and substitute it into y n Get the original accurate short-circuit current component.

9. A method for extracting short-circuit characteristics of a high-proportion distributed power distribution network according to claim 8, characterized in that: After the fault current signal is extracted, multiple modal components containing fault characteristic information are obtained.

Images