Power quality disturbance signal classification method and device considering new energy access

Abstract

The invention provides an electric energy quality disturbance signal classification method and device considering new energy access, and relates to the technical field of data analysis, and the method comprises the steps: obtaining an electric energy quality disturbance signal generated by a new energy power station; variational mode decomposition parameter optimization is carried out based on a sparrow search algorithm, and optimal decomposition parameters of various disturbance signals are determined; decomposing the disturbance signal into a plurality of sub-signals by using an optimal parameter variational mode; and training a classifier by using the constructed feature set, inputting the disturbance signal into the training classifier after training is completed, and outputting the type of the disturbance signal. Through variational mode decomposition parameter optimization based on a sparrow search algorithm, accurate decomposition of signals is realized, much noise caused by decomposition errors is avoided, and the accuracy of signal classification is improved.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and in particular to a method and device for classifying power quality disturbance signals taking into account the access of new energy sources. Background Art

[0002] Due to the output characteristics of new energy power stations, hydropower complementary bases are prone to generate power quality disturbance signals, and there are many distortion components in voltage and current, which can easily cause the reduction of power quality. The essence of power quality disturbance identification is to find suitable feature quantities to distinguish each type of disturbance, construct a feature set, input it into the classifier for training, and finally perform discrimination and classification. Disturbance classification is a prerequisite for improving power quality, and it is also an important basis for dividing problem responsibilities and formulating methods to improve power quality. Summary of the invention

[0003] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0004] To this end, the first objective of the present application is to propose a method for classifying power quality disturbance signals taking into account the access of new energy sources.

[0005] The second object of the present application is to provide a device.

[0006] The third objective of the present application is to provide an electronic device.

[0007] A fourth objective of the present application is to provide a computer-readable storage medium.

[0008] A fifth object of the present application is to provide a computer program product.

[0009] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a method for classifying power quality disturbance signals considering the access of new energy sources, including:

[0010] Obtain disturbance signals of power quality generated by new energy power stations;

[0011] Based on the sparrow search algorithm, the variational mode decomposition parameters are optimized to determine the best decomposition parameters for various disturbance signals;

[0012] Decomposing the disturbance signal into a plurality of sub-signals by using optimal parameter variational mode;

[0013] Extract features from sub-signals and construct them into feature sets;

[0014] The constructed feature set is used to train a classifier, and after the training is completed, the disturbance signal is input into the training classifier and the type of the disturbance signal is output.

[0015] Optionally, performing variational modal decomposition parameter optimization based on a sparrow search algorithm to determine the optimal decomposition parameters of various disturbance signals includes:

[0016] Acquire a J-type disturbance data set, which includes multiple disturbance signals;

[0017] Set the initial parameters of the sparrow algorithm, wherein the initial parameters include a quadratic penalty term α and a decomposition parameter k;

[0018] performing variational mode decomposition on the disturbance signal according to the initial parameters;

[0019] Calculate the minimum envelope information entropy of each position of the J-type disturbance, and calculate the average value of the minimum envelope information entropy;

[0020] In response to the current round being less than the maximum iteration round, updating the parameter to be optimized and re-performing the variational mode decomposition;

[0021] In response to the current round being equal to the maximum iteration round, the optimal parameters and the minimum envelope information entropy are output.

[0022] Optionally, performing variational mode decomposition on the disturbance signal according to the initial parameters includes:

[0023] Decomposing the disturbance signal, the formula is as follows:

[0024]

[0025] Among them, u k Represents each modal signal, w k is the modal center frequency, α is the quadratic penalty term, λ is the Lagrange multiplier, and k is the decomposition parameter;

[0026] initialization in, is the spectrum of the kth sub-signal, is the center frequency of the kth sub-signal;

[0027] The spectrum and center frequency of all modes are updated in multiple rounds of iterations until the objective function converges. The objective function is:

[0028] n is the number of iterations, and ∈ is the acceptable error value.

[0029] Optionally, the features include at least one of the following:

[0030] Multi-scale permutation entropy, maximum value, minimum value, peak value, peak-to-peak value, mean value, average amplitude, root amplitude, variance, standard deviation, effective value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, clearance factor center frequency, wavelet energy ratio, number of zero crossings, energy of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous frequency curve after Hilbert transform, difference in energy distribution of sub-signals decomposed from standard signal, total harmonic distortion, wavelet scale entropy, wavelet singular entropy.

[0031] Optionally, extracting features from the sub-signals and constructing a feature set includes:

[0032] Assume that the time series is {x(i), i = 1, 2, ..., N}, and construct the coarse-grained time series. The formula is as follows:

[0033]

[0034] Where: τ″ is the scale factor. When τ″>1, the original time series is coarse-grained into τ″ coarse-grained series with a length of p=[N / τ″] [·] indicates rounding;

[0035] Calculate the permutation entropy value at each scale factor τ″.

[0036] Optionally, the using the constructed feature set to train the classifier includes:

[0037] Use the histogram algorithm to discretize the eigenvalues ​​into integers, obtain the histogram of the leaf node through the difference between the parent node and the sibling node, and find the best split point;

[0038] A leaf strategy with depth limit is applied based on the histogram, and the node with the largest splitting gain is selected for splitting;

[0039] Perform data sampling and feature sampling and perform classification.

[0040] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a power quality disturbance signal classification device considering the access of new energy sources, including:

[0041] The acquisition module is used to obtain the disturbance signal of the power quality generated by the new energy power station;

[0042] The optimization module is used to optimize the variational mode decomposition parameters based on the sparrow search algorithm to determine the optimal decomposition parameters of various disturbance signals;

[0043] A decomposition module, used for decomposing the disturbance signal into multiple sub-signals by using optimal parameter variational mode decomposition;

[0044] A feature extraction module, used to extract features from sub-signals and construct them into feature sets;

[0045] The classification module is used to train a classifier using the constructed feature set, and after the training is completed, the disturbance signal is input into the training classifier and the type of the disturbance signal is output.

[0046] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0047] The memory stores computer-executable instructions;

[0048] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0049] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer-readable storage medium is stored computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0050] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, which implements any method in the first aspect when executed by a processor.

[0051] The power quality disturbance signal classification method, device, electronic device and storage medium provided in this application considering the access of new energy sources can achieve the following beneficial effects:

[0052] Power quality monitoring and disturbance signal identification: By acquiring the disturbance signal of power quality generated by the new energy power station, the power quality of the power station can be monitored in real time, and power quality problems can be discovered and handled in a timely manner. This is crucial to ensure the stable operation of the power system and improve the efficiency of power utilization.

[0053] Parameter optimization improves decomposition effect: Based on the sparrow search algorithm, the variational mode decomposition parameter optimization can determine the best decomposition parameters for various disturbance signals. This parameter optimization method can improve the effect of variational mode decomposition (VMD), so that the decomposed sub-signals more accurately reflect the characteristics of the original signal. By optimizing the algorithm, the accuracy and efficiency of signal processing can be improved.

[0054] Signal decomposition and sub-signal extraction: Using optimal parameter variational mode decomposition to decompose the disturbance signal into multiple sub-signals can more carefully analyze and identify the different components of power quality disturbances. This decomposition method helps to extract useful information from complex power quality disturbance signals and provides a basis for subsequent feature extraction and classification.

[0055] Feature extraction and feature set construction: Extracting features from sub-signals and constructing feature sets can provide rich feature information for the classification of power quality disturbance signals. These features include kurtosis, entropy, fractal values, etc., which can capture the sharpness, complexity and self-similarity of the signal, thereby effectively distinguishing different types of disturbance signals.

[0056] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0058] Figure 1 A flow chart of a method for classifying power quality disturbance signals taking into account the access of new energy sources provided in an embodiment of the present application;

[0059] Figure 2 A SSA optimized VMD flow chart provided in an embodiment of the present application;

[0060] Figure 3 Schematic diagram of the coarse-graining process under the second scale factor;

[0061] Figure 4 It is a schematic diagram of the histogram algorithm;

[0062] Figure 5 It is the Leaf-Wise growth strategy diagram;

[0063] Figure 6 Schematic diagram of training the differential acceleration model for the histogram;

[0064] Figure 7 It is a schematic diagram of the iteration curve of the sparrow search algorithm;

[0065] Figure 8 Schematic diagram of using CEEMDAN to decompose a composite disturbance signal;

[0066] Fig. 9 is the wavelet packet transform decomposition diagram;

[0067] Fig.10 Optimize VMD decomposition graph for SSA;

[0068] Fig.11 A schematic diagram of the structure of a power quality disturbance signal classification device considering the access of new energy sources provided in an embodiment of the present application. DETAILED DESCRIPTION

[0069] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0070] Due to the output characteristics of new energy power stations, hydropower complementary bases are prone to generate power quality disturbance signals, and there are many distorted components in the voltage and current, which can easily cause a decrease in power quality. The essence of power quality disturbance identification is to find suitable feature quantities to distinguish each type of disturbance, construct a feature set, and input it into the classifier for training, and finally perform discrimination and classification. Disturbance classification is a prerequisite for improving power quality, and it is also an important basis for dividing problem responsibilities and formulating methods to improve power quality. Early power quality disturbance identification relied entirely on manual discrimination, but with the exponential increase in the amount of data, domestic and foreign scholars are urgently needed to develop more efficient computer recognition methods to replace traditional manual recognition methods.

[0071] Disturbance classification includes three processes: disturbance decomposition, feature extraction, and disturbance identification. Feature extraction is to decompose the disturbance waveform using time-frequency analysis tools, and then extract the feature quantity that can realize disturbance classification from the sub-signal. The commonly used signal decomposition tools are FFT, Short-Time Fourier Transform (STFT), WT, EMD, S transform, etc. FFT has good time-frequency analysis capabilities for steady-state signals of power grids, but it has defects such as "side lobes" and "spectrum leakage" in the analysis of composite disturbance signals, and it is difficult to decompose more complex multiple disturbance signals. Although STFT can decompose disturbance signals by combining time domain and frequency domain information, once the window function is selected, the time-frequency resolution capability is determined, and the analysis effect is greatly affected by the window function. Although WT solves the problem of fixed time-frequency resolution of STFT, the selection of wavelet basis and decomposition layer number is usually selected according to the characteristics of the signal, which is greatly affected by human factors. S transform is an extension of STFT and WT, and has good time-frequency analysis capabilities and noise resistance, but it is constrained by the Heisenberg uncertainty principle like STFT. The adaptive ability of Hilbert-Huang transform is quite significant and is not constrained by Heisenberg's uncertainty principle, but the modal aliasing phenomenon is quite serious. The empirical wavelet transform (EWT) is an adaptive time-frequency analysis method developed on the basis of wavelet transform, but the anti-noise performance of the empirical wavelet transform is not strong and needs to be preprocessed according to the frequency domain peak detection, and there are many noise components.

[0072] After signal decomposition and feature extraction, disturbance identification and classification are performed. There are several existing methods for disturbance identification:

[0073] (1) The classification principle and idea of ​​decision tree (DT) is mainly to complete classification based on rule constraints and imitate human logical judgment. DT has small computational complexity, but its shortcomings are also obvious. For example, the feature thresholds of some categories are not easy to determine and are prone to overfitting or local over-optimization.

[0074] (2) Support Vector Machine (SVM) can solve the classification problem in high-dimensional space by classifying two types of samples through a classification surface. It has the advantages of recognizing many types and high classification accuracy, and can obtain the global optimal solution. However, it is difficult to design. When there is not much data available for training, the classification accuracy is not high, and it is relatively difficult to determine the kernel function and parameters.

[0075] (3) Artificial Neural Network (ANN) has a simple structure, high classification accuracy and flexible structure, but its convergence speed and accuracy are greatly affected by structure and noise, and the training time is long.

[0076] In general, the existing power quality disturbance classification method can classify fewer disturbance types, and the classification accuracy needs to be improved. After adding noise, the classification accuracy of the method has a significant decrease. In the actual field environment, double disturbances or even triple disturbances may occur. Therefore, the new disturbance classification algorithm should achieve as many disturbance types as possible and as high a classification accuracy as possible, and at the same time, it should have a certain noise resistance performance.

[0077] The embodiment of the present application provides a method for classifying power quality disturbance signals taking into account the access of new energy sources. Figure 1 The present invention provides a flow chart of a method for classifying power quality disturbance signals taking into account the integration of new energy sources. Figure 1 As shown, the method comprises the following steps:

[0078] Step 101, obtaining a disturbance signal of power quality generated by a new energy power station;

[0079] Step 102, performing variational mode decomposition parameter optimization based on the sparrow search algorithm to determine the optimal decomposition parameters of various disturbance signals;

[0080] Step 103, using optimal parameter variational mode decomposition to decompose the disturbance signal into multiple sub-signals;

[0081] Step 104, extracting features from the sub-signals and constructing a feature set;

[0082] Step 105: train a classifier using the constructed feature set. After the training is completed, input the disturbance signal into the training classifier and output the type of the disturbance signal.

[0083] Optionally, performing variational modal decomposition parameter optimization based on a sparrow search algorithm to determine the optimal decomposition parameters of various disturbance signals includes:

[0084] Acquire a J-type disturbance data set, which includes multiple disturbance signals;

[0085] Set the initial parameters of the sparrow algorithm, wherein the initial parameters include a quadratic penalty term α and a decomposition parameter k;

[0086] performing variational mode decomposition on the disturbance signal according to the initial parameters;

[0087] Calculate the minimum envelope information entropy of each position of the J-type disturbance, and calculate the average value of the minimum envelope information entropy;

[0088] In response to the current round being less than the maximum iteration round, updating the parameter to be optimized and re-performing the variational mode decomposition;

[0089] In response to the current round being equal to the maximum iteration round, the optimal parameters and the minimum envelope information entropy are output.

[0090] Variational mode decomposition (VMD) is an adaptive, completely non-recursive method of modal variation and signal processing. This technology has the advantage of being able to determine the number of modal decompositions. Its adaptability is reflected in determining the number of modal decompositions of a given sequence according to actual conditions. In the subsequent search and solution process, it can adaptively match the optimal center frequency and limited bandwidth of each mode, and can achieve effective separation of intrinsic mode components (IMFs), frequency domain division of signals, and then obtain effective decomposition components of a given signal, and finally obtain the optimal solution to the variational problem. It overcomes the problems of endpoint effects and overlapping of modal components in the EMD method, and has a more solid mathematical theoretical foundation. It can reduce the non-flatness of time series with high complexity and strong nonlinearity, and decompose to obtain relatively flat subsequences containing multiple different frequency scales. It is suitable for non-flat sequences. The core idea of ​​VMD is to construct and solve variational problems.

[0091] VMD regards the sub-signal as a signal with limited bandwidth around the center frequency, so it can be expressed as a constrained variational problem as follows:

[0092]

[0093] Among them, u k Represents each modal signal, w k is the modal center frequency. VMD adds the quadratic penalty term α and the Lagrange multiplier λ, changing the original variational problem to equation (1-2):

[0094]

[0095] Among them, u k Represents each modal signal, w k is the modal center frequency, α is the quadratic penalty term, λ is the Lagrange multiplier, and k is the decomposition parameter; use the alternating direction multiplier method to solve the above equation. Continuously updated Get each mode.

[0096] The basic steps of the VMD algorithm are as follows:

[0097] initialization in, is the spectrum of the kth sub-signal, is the center frequency of the kth sub-signal;

[0098] At the n+1th iteration, for all modes, the frequency spectrum of the mode is updated according to formula (1-3):

[0099]

[0100] The real part of the time-domain analytical signal can be obtained as x k (t). Update the center frequency of the mode using equation (1-4):

[0101]

[0102] The Lagrange multiplier is updated as formula (1-5):

[0103]

[0104] Where τ is the update coefficient.

[0105] The spectrum and center frequency of all modes are updated in multiple rounds of iterations and return to equation (1-2) until the objective function converges. The objective function is:

[0106]

[0107] n is the number of iterations, and ∈ is the acceptable error value.

[0108] The decomposition parameter K and bandwidth parameter α of VMD will seriously affect the classification accuracy of the disturbance signal. If the given K is smaller than the modal number of the actual signal, if a smaller α value is taken, the noise content of some sub-signals will be relatively high, and modal aliasing will occur. To address this problem, the present invention patent uses a sparrow search algorithm to optimize K and α, and find the optimal parameter combination to decompose the disturbance signal, so as to improve the effect of feature extraction, thereby improving the classification accuracy.

[0109] The sparrow search algorithm divides the sparrows in a population into discoverers and followers, and randomly distributes a certain number of alerters. In each iteration, the discoverer updates its position according to formula (1-7).

[0110]

[0111] Where t is the current iteration number, j=1,2…d is the dimension of the parameter to be optimized; iter max is the maximum number of iterations; is the i-th sparrow of dimension j at the t-th iteration; α′ and Q are random numbers, R2 and ST are the alarm value and safety threshold; L is a 1×d all-one matrix.

[0112] If a follower in the population observes that the discoverer has tracked the food, it will compete with the discoverer. The position update formula of the follower is shown in formula (1-8)

[0113]

[0114] Where A + =A T (AA T ) -1 ; Xp is the optimal producer position, X worst is the global worst position; A is a 1×d matrix whose elements are randomly set to 1 and -1.

[0115] Assuming that the number of alerters accounts for 10% to 20% of the total number, the initial position is randomly generated and the position is updated according to formula (1-9).

[0116]

[0117] Where, X best is the global optimal position; β is the step size; f i is the current fitness; f g and f w are the best and worst global fitness, K is a random number between [-1,1]; h is the minimum constant to prevent division by zero. Formula 1-9 completes a round of updates. After several updates, the best fitness and corresponding parameters are found.

[0118] Envelope entropy can reflect the sparse characteristics of the original signal. The patent of this invention uses the minimum average value of the minimum envelope information entropy (MEE) as the fitness function of the sparrow algorithm. If there is more noise and less feature information in the IMF, the sparsity of the component signal is weaker and the envelope entropy value is larger. If there is more feature information contained in the IMF, the sparse characteristics of the signal are stronger and the envelope entropy value is smaller. Envelope Entropy IMF EE The expression is:

[0119]

[0120] The zero-mean signal x(j) (j=1,2,…,N) is transformed by Hilbert to obtain a(j), and a(j) is normalized to obtain e j , IMF EE (k) is the envelope information entropy of the kth modal signal, and the minimum envelope information entropy is obtained by the following formula.

[0121] MEE=min{IMF EE (1),…,IMF EE (k)} (1-11)

[0122] Suppose there are J types of power quality disturbance signals, whose minimum envelope information entropies are MEE(1), MEE(2), MEE(3)...MEE(J) respectively. The global fitness function is the minimum average value of the minimum envelope information entropy of J types of disturbance signals.

[0123]

[0124] Through SSA, the parameters obtained by VMD are globally optimized to obtain the optimal parameters corresponding to the minimum fitness function value. The specific steps are as follows: Figure 2 shown.

[0125] Optionally, the features include at least one of the following:

[0126] Multi-scale permutation entropy, maximum value, minimum value, peak value, peak-to-peak value, mean value, average amplitude, root amplitude, variance, standard deviation, effective value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, clearance factor center frequency, wavelet energy ratio, number of zero crossings, energy of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous frequency curve after Hilbert transform, difference in energy distribution of sub-signals decomposed from standard signal, total harmonic distortion, wavelet scale entropy, wavelet singular entropy.

[0127] Optionally, extracting features from the sub-signals and constructing a feature set includes:

[0128] Assume that the time series is {x(i), i=1,2,…,N}, and construct the coarse-grained time series. The formula is as follows:

[0129]

[0130] Where: τ″ is the scale factor. When τ″>1, the original time series is coarse-grained into τ″ coarse-grained series with a length of p=[N / τ″] [·] indicates rounding;

[0131] Calculate the permutation entropy value at each scale factor τ″.

[0132] Composite Multiscale Permutation Entropy (CMPE) improves the single coarse-graining of multiscale permutation entropy to composite coarse-graining. The specific calculation steps are as follows:

[0133] The sequence {x(i), i = 1, 2, ..., N} is defined as a coarse-grained sequence by Right now:

[0134]

[0135] represents the kth coarse-grained sequence under the scale factor τ′, and j represents The jth point of .

[0136] For τ′, calculate each coarse-grained sequence under this scale factor PE, and then average the τ′ PE values ​​to get the value of CMPE under the scale factor τ′, that is:

[0137]

[0138] Where m is the embedding dimension and λ′ is the delay time.

[0139] Figure 3 is a schematic diagram of the coarse-graining process under the second scale factor, as shown in Figure 3 As shown, the original signal time series is divided into two coarse-grained time series.

[0140] Optionally, the using the constructed feature set to train the classifier includes:

[0141] Use the histogram algorithm to discretize the eigenvalues ​​into integers, obtain the histogram of the leaf node through the difference between the parent node and the sibling node, and find the best split point;

[0142] A leaf strategy with depth limit is applied based on the histogram, and the node with the largest splitting gain is selected for splitting;

[0143] Perform data sampling and feature sampling and perform classification.

[0144] In this embodiment, power quality disturbance classification is performed based on optimized variational modal decomposition-composite multi-scale permutation entropy-LightGBM. Traditional boosting algorithms need to traverse the training data multiple times in each iterative operation, which increases the time for model training and is therefore difficult to adapt to power systems with huge amounts of data. LightGBM performs better than GBDT when faced with huge amounts of data. It is a high-speed, distributed, and high-performance gradient boosting framework that is more suitable for classification and identification of power quality disturbances.

[0145] Boosting needs to scan all features when selecting the optimal split point, which takes a lot of time on a data set with a large feature dimension. LightGBM uses a histogram algorithm to discretize the feature values ​​into integers, construct a histogram, and find the best split point. The histogram algorithm is as follows Figure 4 As shown in the figure. Based on the histogram, the hierarchical strategy of GBDT is not used as the growth method of the decision tree. The leaf strategy with depth restriction is applied to select the node with the largest split gain for splitting. The Leaf-Wise growth strategy is as follows Figure 5 In addition to the leaf-by-leaf tree growth strategy, another optimization of LightGBM is to obtain the histogram of the leaf node through the difference between the parent node and the sibling node. The histogram difference accelerates the training of the model. Figure 6 shown.

[0146] In the classification process of LightGBM, data sampling and feature sampling are performed simultaneously. High-dimensional data is usually sparse and mutually exclusive. Unlike conventional feature sampling, the feature bundling algorithm fuses mutually exclusive features to reduce the dimension of features and simplifies it into a graph coloring problem. A greedy strategy is used in the solution process. The feature bundling algorithm does not scan all features, which improves the efficiency of the algorithm while ensuring accuracy.

[0147] In a possible embodiment, assuming that there are J categories of single and composite disturbances that need to be classified, first collect J types of disturbance signals to construct a data set, use the sparrow search algorithm to optimize the parameters of variational mode decomposition, determine the number of decomposition layers K and the bandwidth parameter α, and the fitness function is as described above. After the optimal parameters are determined, the disturbance signal is decomposed into several sub-signals, and several time domain and frequency domain features including composite multi-scale permutation entropy are extracted from the sub-signals to construct a feature set. Finally, the feature set is used to train the classifier LightGBM, and the disturbance signal is classified after the training is completed.

[0148] In a possible embodiment, a simulation experiment is conducted to simulate and generate 7 single disturbances, including standard signal (C0), temporary rise (C1), interruption (C2), temporary sag (C3), harmonic (C4), transient oscillation (C5), transient pulse (C6), and voltage flicker (C7); 7 double disturbances, including interruption plus harmonic (C2+C4), interruption plus transient pulse (C2+C6), harmonic plus transient oscillation (C4+C5), harmonic plus transient pulse (C4+C6), harmonic plus flicker (C4+C7), transient oscillation plus flicker (C5+C7), and transient pulse plus flicker (C6+C7); and 2 triple disturbances, including temporary sag plus transient oscillation plus voltage flicker (C3+C5+C7), and transient oscillation plus transient pulse plus voltage flicker (C5+C6+C7). 600 experimental data are randomly generated for each type of disturbance, with a sampling frequency of 6.4KHz and a sampling length of 2048 sampling points. The experiment adopted ten-fold cross validation.

[0149] Usually, the collected field voltage and current signals are easily mixed with noise, which may contain both wide-range and narrow-range frequency components. Therefore, the value of parameter τ is fixed to 0. α and K are optimized globally according to the algorithm mentioned above. The number of SSA groups is set to 40, and the maximum number of iterations is set to 80. The optimal parameter combination is [K, α] = [10, 237]. The sparrow search algorithm iteration curve is shown in Figure 7 shown.

[0150] The parameters of the optimization model are analyzed based on the power quality composite disturbance example. The flicker + high-order harmonic composite disturbance signal is constructed: the 3rd, 5th, 7th, 9th, 11th, and 13th harmonic signals with an amplitude of 0.2 and 50dB Gaussian white noise. The flicker disturbance part is as shown in formula (1-16), where α = 0.45 and β = 0.2.

[0151] u(t)=[1+αsin(βwt)]sin(wt) (1-16)

[0152] Figure 8 This is a schematic diagram of using CEEMDAN to decompose a composite disturbance signal. 13 modes are decomposed. It can be seen from the figure that there is a serious "modal aliasing" phenomenon. It is difficult to distinguish between harmonic signals and flicker signals due to the interference of noise signals.

[0153] The disturbance signal is decomposed by wavelet packet transform. After experimental comparison, the sub-signal obtained by three-layer decomposition of the "db4" mother wavelet can more accurately decompose the fundamental wave and flicker signal. Fig. 9 As shown, IMF1 is a flicker signal, but the waveform is distorted, and the other harmonic signals also have "modal aliasing" phenomenon, resulting in serious waveform distortion. The wavelet packet transform is also greatly affected by noise.

[0154] The constructed signal is decomposed by VMD using the optimal parameter combination, such as Fig.10 As shown in the figure, IMF1 is a flicker signal, IMF2-IMF7 are harmonic signals with an amplitude of 0.2, and IMF8-IMF10 are decomposed false components. The decomposed sub-signals do not have the "modal aliasing" phenomenon, and the amplitudes of the sub-signals are also relatively accurate. It can be seen that using the sparrow search algorithm to adaptively optimize the decomposition parameters of VMD can achieve better decomposition effects.

[0155] After the constructed 17 types of signals are subjected to optimal parameter VMD decomposition, the features including the above CMPE are extracted to form a feature set. The embedding dimension of CMPE is selected as 6, and the scale factor is set to 12. Due to the high dimension of CMPE, PCA is used to reduce the dimension and extract 3 principal components to reduce the complexity of the features. LightGBM is selected for classification experiments. In order to prove the classification effect of the method proposed in the present invention in a noisy environment, disturbance signals in 30dB, 40dB, and 50dB noise environments are constructed. The classification results are shown in Table 1.1. As can be seen from Table 1.1, the method proposed in the present invention can achieve an accuracy of up to 98.627% without noise, and even in triple disturbances, there is a high recognition accuracy. In the three different degrees of noise environments constructed, the recognition accuracy is only slightly reduced. In the 30dB noise environment, the overall accuracy can still reach more than 97%.

[0156] To verify the advantages of LightGBM classifier in classification accuracy and classification speed. The commonly used machine learning classifiers XGboost, CATboost and random forest are used to compare the classification accuracy and classification time in different noise environments as shown in Table 1.2 and Table 1.3. It can be seen from Table 1.2 and Table 1.3 that under different noise backgrounds, LightGBM has higher classification accuracy and lower time consumption than other classifiers in most noise environments. Random forest also has lower classification time consumption due to its simple structure, but its classification accuracy is lower. The classification time consumption of CATboost is significantly higher than that of several other classifiers.

[0157] Table 1.1 Classification accuracy of the method proposed in this patent under different noise environments

[0158]

[0159] Table 1.2 Comparison of classification accuracy of different classifiers

[0160]

[0161] Table 1.3 Comparison of classification time of different classifiers

[0162]

[0163] In order to implement the above-mentioned embodiment, the present application also proposes a power quality disturbance signal classification device considering the access of new energy. Fig.11 A schematic diagram of the structure of a power quality disturbance signal classification device considering the access of new energy sources provided in an embodiment of the present application. Fig.11 As shown, the device comprises:

[0164] The acquisition module 1110 is used to obtain the disturbance signal of the power quality generated by the new energy power station;

[0165] An optimization module 1120 is used to optimize the variational mode decomposition parameters based on the sparrow search algorithm to determine the best decomposition parameters of various disturbance signals;

[0166] A decomposition module 1130, configured to decompose the disturbance signal into a plurality of sub-signals using optimal parameter variational mode decomposition;

[0167] A feature extraction module 1140 is used to extract features from sub-signals and construct them into feature sets;

[0168] The classification module 1150 is used to train a classifier using the constructed feature set, and after the training is completed, the disturbance signal is input into the training classifier and the type of the disturbance signal is output.

[0169] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0170] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0171] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0172] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0173] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0174] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0175] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0176] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0177] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0178] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0179] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0180] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0181] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0182] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for classifying power quality disturbance signals considering the access of new energy sources, characterized in that: The following steps are involved: Obtain disturbance signals of power quality generated by new energy power stations; Based on the sparrow search algorithm, the variational mode decomposition parameters are optimized to determine the best decomposition parameters for various disturbance signals; Decomposing the disturbance signal into a plurality of sub-signals by using optimal parameter variational mode; Extract features from sub-signals and construct them into feature sets; The constructed feature set is used to train a classifier, and after the training is completed, the disturbance signal is input into the training classifier and the type of the disturbance signal is output.

2. The method according to claim 1, characterized in that The variational modal decomposition parameter optimization based on the sparrow search algorithm is performed to determine the optimal decomposition parameters of various disturbance signals, including: Acquire a J-type disturbance data set, which includes multiple disturbance signals; Set the initial parameters of the sparrow algorithm, wherein the initial parameters include a quadratic penalty term α and a decomposition parameter k; performing variational mode decomposition on the disturbance signal according to the initial parameters; Calculate the minimum envelope information entropy of each position of the J-type disturbance, and calculate the average value of the minimum envelope information entropy; In response to the current round being less than the maximum iteration round, updating the parameter to be optimized and re-performing the variational mode decomposition; In response to the current round being equal to the maximum iteration round, the optimal parameters and the minimum envelope information entropy are output.

3. The method according to claim 2, characterized in that The performing variational mode decomposition on the disturbance signal according to the initial parameters comprises: Decomposing the disturbance signal, the formula is as follows: Among them, u k Represents each modal signal, w k is the modal center frequency, α is the quadratic penalty term, λ is the Lagrange multiplier, and k is the decomposition parameter; initialization in, is the spectrum of the kth sub-signal, is the center frequency of the kth sub-signal; The spectrum and center frequency of all modes are updated in multiple rounds of iterations until the objective function converges. The objective function is: n is the number of iterations, and ∈ is the acceptable error value.

4. The method according to claim 3, characterized in that: The features include at least one of the following: Multi-scale permutation entropy, maximum value, minimum value, peak value, peak-to-peak value, mean value, average amplitude, root amplitude, variance, standard deviation, effective value, kurtosis, skewness, waveform factor, peak factor, pulse factor, margin factor, clearance factor center frequency, wavelet energy ratio, number of zero crossings, energy of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous amplitude curve after Hilbert transform, standard deviation of instantaneous frequency curve after Hilbert transform, difference in energy distribution of sub-signals decomposed from standard signal, total harmonic distortion, wavelet scale entropy, wavelet singular entropy.

5. The method according to claim 4, characterized in that The extracting features from the sub-signals and constructing a feature set includes: Assume that the time series is {x(i), i = 1, 2, ..., N}, and construct the coarse-grained time series. The formula is as follows: Where: τ″ is the scale factor. When τ″>1, the original time series is coarse-grained into τ″ coarse-grained series with a length of p=[N / τ″] [·] indicates rounding; Calculate the permutation entropy value at each scale factor τ″.

6. The method according to claim 5, characterized in that The method of training a classifier using the constructed feature set includes: Use the histogram algorithm to discretize the eigenvalues ​​into integers, obtain the histogram of the leaf node through the difference between the parent node and the sibling node, and find the best split point; A leaf strategy with depth limit is applied based on the histogram, and the node with the largest splitting gain is selected for splitting; Perform data sampling and feature sampling and perform classification.

7. A power quality disturbance signal classification device considering the access of new energy, characterized in that: include: The acquisition module is used to obtain the disturbance signal of the power quality generated by the new energy power station; The optimization module is used to optimize the variational mode decomposition parameters based on the sparrow search algorithm to determine the optimal decomposition parameters of various disturbance signals; A decomposition module, used for decomposing the disturbance signal into multiple sub-signals by using optimal parameter variational mode; A feature extraction module, used to extract features from sub-signals and construct them into feature sets; The classification module is used to train a classifier using the constructed feature set, and after the training is completed, the disturbance signal is input into the training classifier and the type of the disturbance signal is output.

8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images