Harmonic detection method, device and equipment

Abstract

The invention provides a harmonic detection method, device and equipment. The method comprises the following steps: acquiring a to-be-decomposed voltage signal; acquiring characteristic data of the to-be-decomposed voltage signal according to the to-be-decomposed voltage signal; inputting the feature data into a pre-trained parameter prediction model to obtain a decomposition number and a penalty factor; wherein the parameter prediction model is a model which is obtained by training based on a machine learning algorithm in advance and is used for predicting decomposition parameters according to voltage signal characteristics; and according to the decomposition number and the penalty factor, processing the to-be-decomposed voltage signal by adopting a variational mode decomposition (VMD) algorithm to obtain a plurality of intrinsic mode function (IMF) components corresponding to the to-be-decomposed voltage signal. The method is used for achieving the effect of improving the harmonic detection efficiency of the micro-grid.

Description

Technical Field

[0001] This application relates to the field of power grids, and in particular to a harmonic detection method, apparatus and equipment. Background Technology

[0002] Harmonic detection in microgrids is crucial for ensuring their normal operation. With the rapid development of energy transition and distributed power generation, hybrid AC / DC microgrids, leveraging their combined AC and DC power supply advantages, have become the mainstream form for achieving efficient energy utilization. However, the large-scale integration of nonlinear loads such as photovoltaic inverters, energy storage converters, and electric vehicle charging stations has led to higher harmonic content and more complex characteristics in microgrids, placing higher demands on harmonic detection technology.

[0003] Currently, commonly used harmonic detection algorithms include Fast Fourier Transform, Wavelet Transform, and Variational Mode Decomposition (VMD). Among these, VMD, due to its ability to adaptively partition the original signal into Intrinsic Mode Function (IMF) components, can largely avoid mode aliasing and can be applied to fault detection in microgrids with significant harmonic pollution. When using the VMD algorithm for harmonic detection, the number of decompositions and the penalty factor are crucial to the effectiveness of harmonic detection.

[0004] However, existing commonly used algorithms (such as grid search algorithms) often take a long time to calculate the number of decompositions and the penalty factor, resulting in low computational efficiency and failing to meet the real-time requirements for harmonic detection in microgrids. Summary of the Invention

[0005] This application provides a harmonic detection method, apparatus, and device to improve the efficiency of harmonic detection in microgrids.

[0006] In a first aspect, embodiments of this application provide a harmonic detection method, including:

[0007] Acquire the voltage signal to be decomposed;

[0008] Based on the voltage signal to be decomposed, obtain the feature data of the voltage signal to be decomposed;

[0009] The feature data is input into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor; wherein, the parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm.

[0010] Based on the number of decompositions and the penalty factor, the variational mode decomposition (VMD) algorithm is used to process the voltage signal to be decomposed, thereby obtaining multiple intrinsic mode function (IMF) components corresponding to the voltage signal to be decomposed.

[0011] In one possible implementation, before inputting the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor, the method further includes:

[0012] A sample set is obtained, which includes: feature data corresponding to multiple different voltage signals, the number of calibrated decompositions, and penalty factors; the multiple different voltage signals are collected when the microgrid is not connected to nonlinear devices and when it is connected to different nonlinear devices;

[0013] The parameter prediction model is obtained by training the pre-constructed initial support vector regression model using the sample set.

[0014] In one possible implementation, the kernel function in the initial support vector regression model is a hybrid kernel function obtained by combining the radial basis function and the sigmoid kernel function.

[0015] In one possible implementation, obtaining the sample set includes:

[0016] Signals were acquired from the AC bus of the microgrid under two conditions: with no nonlinear devices connected and with different nonlinear devices connected. Multiple voltage signals were obtained.

[0017] For each voltage signal, characteristic data of the voltage signal are obtained based on the voltage signal;

[0018] Based on the grid search algorithm, the number of decompositions and the penalty factor corresponding to the voltage signal are determined;

[0019] The sample set is constructed based on the feature data of each voltage signal, the number of decompositions corresponding to each voltage signal, and the penalty factor.

[0020] In one possible implementation, the feature data includes at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

[0021] In one possible implementation, the method further includes:

[0022] The loss function selected during model training for:

[0023]

[0024] Where N represents the number of training samples, w iLet y represent the weight of the i-th sample. i This represents the true label of the i-th sample. Let represent the predicted label of the i-th sample, and ε represent the threshold.

[0025] In one possible implementation, the method further includes:

[0026] Based on the voltage signal to be decomposed and multiple IMF components, the energy capture rate PER is calculated, and the linear correlation between any two IMF components is calculated.

[0027] If the PER is less than the first preset value, then the number of decompositions is incremented by 1;

[0028] If the maximum value of the linear correlation between any two IMF components is greater than the second preset value, then the number of decompositions is reduced by 1.

[0029] Based on the updated number of decompositions, the VMD algorithm is used to process the voltage signal to be decomposed, and new IMF components corresponding to the voltage signal to be decomposed are obtained.

[0030] In one possible implementation, the method further includes:

[0031] The short-time energy change rate ΔE of the voltage signal to be decomposed is calculated in real time.

[0032] If ΔE is not 0, then according to ΔE, the formula α is used. new =α pre The penalty factor is updated using (1 + 0.1 tanh(ΔE / 0.2)), where α new Let α represent the updated penalty factor. pre Indicates the penalty factor before the update;

[0033] Based on the updated penalty factor, the VMD algorithm is used to process the voltage signal to be decomposed, and new IMF components corresponding to the voltage signal to be decomposed are obtained.

[0034] In one possible implementation, the method further includes:

[0035] After each decomposition of the voltage signal to be decomposed using the VMD algorithm to obtain multiple IMF components, the energy capture rate PER, modal redundancy MRI, and signal-to-noise ratio SNR corresponding to the multiple IMF components are calculated.

[0036] Based on the PER, the MRI, and the SNR, the decomposition quality evaluation index Q is calculated using the formula Q = m1PER + m2(1-MRI) + m3SNR, where m1, m2, and m3 are the weighting coefficients of the PER, the MRI, and the SNR, respectively; Q is used to indicate the quality of the IMF components obtained from the decomposition.

[0037] If Q exceeds a preset range, the parameter prediction model is iteratively trained.

[0038] In one possible implementation, the method further includes:

[0039] If the number of times Q exceeds the preset range is greater than the preset number, then incremental learning is used to update the parameter prediction model.

[0040] Secondly, embodiments of this application provide a harmonic detection device, comprising:

[0041] The first acquisition unit is used to acquire the voltage signal to be decomposed;

[0042] The first processing unit is used to obtain the feature data of the voltage signal to be decomposed based on the voltage signal to be decomposed.

[0043] The second processing unit is used to input the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor; wherein, the parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm.

[0044] The third processing unit is used to process the voltage signal to be decomposed using the variational mode decomposition (VMD) algorithm according to the number of decompositions and the penalty factor, so as to obtain multiple intrinsic mode function (IMF) components corresponding to the voltage signal to be decomposed.

[0045] In one possible implementation, the harmonic detection device further includes:

[0046] The second acquisition unit is used to acquire a sample set, which includes: feature data corresponding to multiple different voltage signals, the calibrated number of decompositions, and a penalty factor; the multiple different voltage signals are collected when the microgrid is not connected to nonlinear devices and when it is connected to different nonlinear devices;

[0047] The training unit is used to train the pre-constructed initial support vector regression model using the sample set to obtain the parameter prediction model.

[0048] In one possible implementation, the kernel function in the initial support vector regression model of the training unit is a hybrid kernel function obtained by combining the radial basis function and the sigmoid kernel function.

[0049] In one possible implementation, the second acquisition unit is specifically used for:

[0050] Signals were acquired from the AC bus of the microgrid under two conditions: with no nonlinear devices connected and with different nonlinear devices connected. Multiple voltage signals were obtained.

[0051] For each voltage signal, characteristic data of the voltage signal are obtained based on the voltage signal;

[0052] Based on the grid search algorithm, the number of decompositions and the penalty factor corresponding to the voltage signal are determined;

[0053] The sample set is constructed based on the feature data of each voltage signal, the number of decompositions corresponding to each voltage signal, and the penalty factor.

[0054] In one possible implementation, the feature data includes at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

[0055] In one possible implementation, the training unit selects a loss function during model training. for:

[0056]

[0057] Where N represents the number of training samples, w i Let y represent the weight of the i-th sample. i This represents the true label of the i-th sample. Let represent the predicted label of the i-th sample, and ε represent the threshold.

[0058] In one possible implementation, the harmonic detection device further includes:

[0059] The first update unit is used to calculate the energy capture rate (PER) based on the voltage signal to be decomposed and multiple IMF components, and to calculate the linear correlation between any two IMF components; if the PER is less than a first preset value, the number of decompositions is incremented by 1; if the maximum value of the linear correlation between any two IMF components is greater than a second preset value, the number of decompositions is decremented by 1; based on the updated number of decompositions, the VMD algorithm is used to process the voltage signal to be decomposed to obtain new multiple IMF components corresponding to the voltage signal to be decomposed.

[0060] In one possible implementation, the harmonic detection device further includes:

[0061] The second update unit is used to calculate the short-time energy change rate ΔE of the voltage signal to be decomposed in real time.

[0062] If ΔE is not 0, then according to ΔE, the formula α is used. new =α pre The penalty factor is updated using (1 + 0.1 tanh(ΔE / 0.2)), where α new Let α represent the updated penalty factor. pre The penalty factor before the update is represented; based on the updated penalty factor, the VMD algorithm is used to process the voltage signal to be decomposed to obtain multiple new IMF components corresponding to the voltage signal to be decomposed.

[0063] In one possible implementation, the harmonic detection device further includes:

[0064] The verification unit is used to calculate the energy capture rate (PER), modal redundancy (MRI), and signal-to-noise ratio (SNR) of the multiple IMF components after each decomposition of the voltage signal to be decomposed using the VMD algorithm.

[0065] Based on the PER, the MRI, and the SNR, the decomposition quality evaluation index Q is calculated using the formula Q = m1PER + m2(1-MRI) + m4SNR, where m1, m2, and m3 are the weighting coefficients of the PER, the MRI, and the SNR, respectively; and Q is used to indicate the quality of the IMF components obtained from the decomposition.

[0066] If Q exceeds a preset range, the parameter prediction model is iteratively trained.

[0067] In one possible implementation, the harmonic detection device further includes:

[0068] The learning unit is used to update the parameter prediction model by incremental learning if the number of times Q exceeds a preset range is greater than a preset number.

[0069] Thirdly, embodiments of this application provide a computer device, including: a memory and a processor;

[0070] The memory stores computer-executed instructions;

[0071] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0072] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0073] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0074] The harmonic detection method, apparatus, and device provided in this application first acquire the voltage signal to be decomposed, then acquire the feature data of the voltage signal to be decomposed based on the voltage signal to be decomposed, and then input the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor. The parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm. Finally, based on the number of decompositions and the penalty factor, the variational mode decomposition (VMD) algorithm is used to process the voltage signal to be decomposed to obtain multiple intrinsic mode functions (IMF) components corresponding to the voltage signal to be decomposed. Based on the features of the voltage signal and the pre-trained model, the number of decompositions and the penalty factor can be obtained quickly and with high accuracy, which improves the efficiency of harmonic detection based on VMD, thereby improving the efficiency of harmonic detection in microgrids. Attached Figure Description

[0075] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0076] Figure 1 A schematic flowchart of the harmonic detection method provided in Embodiment 1 of this application;

[0077] Figure 2 A schematic diagram illustrating the principle of the specific harmonic detection method provided in Embodiment 3 of this application;

[0078] Figure 3 This is a schematic diagram of the structure of the harmonic detection device provided in Embodiment 4 of this application;

[0079] Figure 4 This is a schematic diagram of the structure of the harmonic detection device provided in Embodiment 5 of this application;

[0080] Figure 5 A schematic diagram of the structure of the computer device provided in this application.

[0081] The accompanying drawings have illustrated specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0082] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0083] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0084] Figure 1 A schematic flowchart of the harmonic detection method provided in Embodiment 1 of this application is shown below. Figure 1 As shown, the method includes:

[0085] S101. Obtain the voltage signal to be decomposed.

[0086] In this step, the voltage signal to be decomposed refers to the voltage signal for which harmonic detection is required. In practical applications, the voltage signal to be decomposed can be a voltage signal collected from the AC bus of a microgrid.

[0087] S102. Based on the voltage signal to be decomposed, obtain the characteristic data of the voltage signal to be decomposed.

[0088] In this step, feature data of the voltage signal will be extracted from the voltage signal to be decomposed. Feature data refers to data used to characterize the properties of the voltage signal, such as time-domain features, frequency-domain features, time-frequency features, and nonlinear features.

[0089] Optionally, the feature data can be at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

[0090] In one specific implementation, the characteristic data of the voltage signal to be decomposed can be obtained from the voltage signal to be decomposed using the following method:

[0091] Step 1: Convert the voltage signal to be decomposed to the dq coordinate system to obtain the q component and d component of the voltage signal to be decomposed;

[0092] Step 2: Based on the q-component and d-component of the voltage signal to be decomposed, calculate the characteristic data of the voltage signal to be decomposed.

[0093] S103. Input the feature data into the pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor.

[0094] In this step, the feature data needs to be input into the pre-trained parameter prediction model to obtain the number of decompositions K and the penalty factor α, which are used to provide the parameter basis for harmonic detection based on the Variational Mode Decomposition (VMD) algorithm.

[0095] The parameter prediction model is a model that predicts decomposition parameters based on voltage signal characteristics, pre-trained using a machine learning algorithm. This machine learning algorithm can be, for example, a decision tree algorithm, a random forest algorithm, or a support vector machine. This application does not impose any restrictions on the selection of the machine learning algorithm.

[0096] It should be understood that the selection of the number of decompositions K and the penalty factor α determines the quality of the IMF components obtained from the decomposition of the voltage signal to be decomposed, thus affecting the accuracy of harmonic detection in the microgrid. This will be explained in detail below in conjunction with the essence of the VMD algorithm:

[0097] The VMD algorithm is an adaptive signal processing method whose core idea is to decompose a complex original signal into several intrinsic mode functions (IMFs) with limited bandwidth. In this algorithm, if the pre-selected number of decompositions K is too large, the center frequencies of adjacent modes will be close, indicating over-decomposition of the signal and difficulty in identifying effective components. If the pre-selected number of decompositions K is too small, insufficient component decomposition will lead to bandwidth overlap and unstable decomposition of the original signal, resulting in mode aliasing. Furthermore, if the pre-selected penalty factor is inappropriate, it will cause the center frequencies of the IMF components to shift and become distorted.

[0098] S104. Based on the number of decompositions and the penalty factor, the variational mode decomposition algorithm is used to process the voltage signal to be decomposed, and multiple intrinsic mode functions (IMF) components corresponding to the voltage signal to be decomposed are obtained.

[0099] In this step, based on the determined number k and penalty factor α, the VMD algorithm is used to process the voltage signal to be decomposed, and k IMF components corresponding to the voltage signal to be decomposed are obtained, thereby completing the decomposition of the voltage signal to be decomposed and realizing the harmonic detection of the microgrid.

[0100] The harmonic detection method, apparatus, and device provided in this embodiment first acquire the voltage signal to be decomposed, then acquire the feature data of the voltage signal to be decomposed based on the voltage signal, and then input the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor. The parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm. Finally, based on the number of decompositions and the penalty factor, the variational mode decomposition (VMD) algorithm is used to process the voltage signal to be decomposed to obtain multiple intrinsic mode functions (IMF) components corresponding to the voltage signal to be decomposed. Based on the features of the voltage signal and the pre-trained model, the number of decompositions and the penalty factor can be obtained quickly and with high accuracy, replacing existing algorithms such as empirical calculation or grid search, thus improving the efficiency of harmonic detection based on VMD and thereby improving the efficiency of harmonic detection in microgrids.

[0101] Furthermore, in one possible implementation, the above parameter prediction model is a model that predicts decomposition parameters based on voltage signal features, pre-trained using a support vector regression classification algorithm. This model can be trained using the following method:

[0102] Step 1: Obtain the sample set, which includes: feature data corresponding to multiple different voltage signals, the number of calibrated decompositions, and penalty factors. The multiple different voltage signals are collected when the microgrid is not connected to nonlinear devices and when it is connected to different nonlinear devices.

[0103] In this step, the obtained sample set includes multiple samples, and each sample includes a set of feature data corresponding to a different voltage signal, the number of calibrated decompositions, and a penalty factor.

[0104] Optionally, the sample set can be obtained in the following ways:

[0105] Step 1.1: Acquire signals from the AC bus of the microgrid under the conditions of no nonlinear equipment connected and with different nonlinear equipment connected, and obtain multiple voltage signals.

[0106] In this step, in order to obtain voltage signals with different harmonic characteristics, it is necessary to collect signals from the AC bus of the microgrid under the condition that the microgrid is not connected to nonlinear devices and under the condition that the microgrid is connected to different nonlinear devices, so as to obtain the voltage signals of the AC bus when the microgrid is not connected to nonlinear devices and the voltage signals of the AC bus when the microgrid is connected to different nonlinear devices, thus obtaining multiple voltage signals.

[0107] Step 1.2: For each voltage signal, obtain the characteristic data of the voltage signal based on the voltage signal.

[0108] In this step, feature data will be extracted for each voltage signal acquired.

[0109] Optionally, the feature data may include at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

[0110] It should be noted that the type of feature data extracted from the voltage signal when the user obtains the sample set here corresponds to the type of feature data extracted from the voltage signal to be decomposed mentioned earlier.

[0111] Step 1.3: Based on the grid search algorithm, determine the number of decompositions and the penalty factor corresponding to the voltage signal.

[0112] In this step, for each voltage signal, a grid search algorithm is used to systematically traverse the preset parameter combinations and select the optimal solution by combining evaluation indicators, so as to determine the optimal number of decompositions and the optimal penalty factor when using the VDM algorithm to decompose the voltage signal.

[0113] Step 1.4: Construct a sample set based on the feature data of each voltage signal, the number of decompositions corresponding to each voltage signal, and the penalty factor.

[0114] In this step, for each voltage signal, the feature data of that voltage signal, along with the number of decompositions and the penalty factor corresponding to that voltage signal, are taken as a sample, thus obtaining a sample set. The number of decompositions and the penalty factor in each sample are labeled as the label of that sample.

[0115] This method first collects different voltage signals when the microgrid is in different operating states, then extracts features from each voltage signal, and determines the optimal number of decompositions and penalty factors. This successfully constructs a sample set that includes feature data corresponding to multiple different voltage signals, a calibrated number of decompositions, and penalty factors. This ensures the diversity of features contained in the sample and effectively guarantees the reliability of the parameter prediction model trained based on this sample set.

[0116] Step 2: Use the sample set to train the pre-built initial support vector regression model to obtain the parameter prediction model.

[0117] In this step, a pre-built initial support vector regression model needs to be trained using a sample set. The mapping relationship between voltage signal features and VMD decomposition parameters (number of decompositions and penalty factor) is fitted using historical data. During training, the kernel matrix is ​​calculated based on the kernel function, and the model parameters are optimized based on a pre-selected loss function to obtain the parameter prediction model. The parameters of the support vector regression model include the parameters in the kernel function of the model.

[0118] The kernel function in the initial support vector regression model can be a commonly used kernel parameter for support vector models, such as the radial basis function kernel (RBF), linear kernel function, and polynomial kernel function. This application does not restrict the choice of kernel function for the model.

[0119] The loss function used for training can be a commonly used loss function for training support vector models, such as hinge loss, logistic loss, exponential loss, etc. This application does not impose any restrictions on the selection of the loss function.

[0120] Optionally, the loss function selected during model training. It can be:

[0121]

[0122] Where N represents the number of training samples, w i The weight of the i-th sample is y, which can be determined in advance based on the sample's entropy. i This represents the true label of the i-th sample. Let represent the predicted label of the i-th sample, and ε represent the threshold.

[0123] Optionally, the kernel function in the initial support vector regression model is a hybrid kernel function obtained by combining the radial basis function (RBF) and the sigmoid kernel function.

[0124] For example, the hybrid kernel function K(X) i ,X j The expression for ) can be represented by the following formula:

[0125] K(X i ,X j )=λ·K RBF (X i ,X j )+(1-λ)K sigmoid (X i ,X j )

[0126] Among them, K RBF (X i ,X j ) represents the radial basis function kernel function, K sigmoid (X i ,X j ) represents the Sigmoid kernel function, and λ represents the weight parameters of the radial basis function kernel.

[0127] The radial basis function kernel function can be expressed by the following formula:

[0128] K RBF (X i ,X j )=tanh(γs ij +c)

[0129] s ij =x i1 x j1 +x i2 x j2 +…+x in x jn

[0130] The Sigmoid kernel function can be expressed by the following formula:

[0131]

[0132] In the above formula, s ij d represents the inner product between the i-th sample and the j-th sample in the sample set. ij Let represent the Euclidean distance between the i-th and j-th samples in the sample set, γ represent the kernel parameter, and c represent the offset.

[0133] Optionally, in practical applications, before training the model using the sample set, each group of feature data in the sample set can be standardized. For example, for any feature data, the following formula can be used for standardization:

[0134]

[0135] Where x represents the feature data, μ represents the mean of the feature data of this type in the sample set, and σ represents the standard deviation of the feature data of this type in the sample set.

[0136] This implementation method first collects multiple different voltage signals with different harmonic characteristics when the microgrid is without nonlinear equipment and with different nonlinear equipment. Based on these multiple different voltage signals, a sample set is obtained, including feature data corresponding to the multiple different voltage signals, the calibrated number of decompositions, and the penalty factor. Then, the sample set is used to train a pre-constructed initial support vector regression model to obtain a parameter prediction model. The sample set covers the voltage harmonic characteristics of the microgrid under various operating conditions, making the model adaptable to the parameter prediction needs of the microgrid in multiple scenarios. This effectively improves the generalization ability of the parameter prediction model and ensures that the parameters obtained based on the training set provide a foundation for the harmonic detection method based on the parameter prediction model, guaranteeing the accuracy of the number of decompositions and the penalty factor determined by the parameter prediction model.

[0137] Embodiment 2 of this application provides a harmonic detection method. Based on the above embodiments, the method provided in this application embodiment further includes the following:

[0138] The first aspect includes a method for dynamically adjusting the number of decompositions and dynamically updating the multiple IMF components obtained from the decomposition based on the adjustment results:

[0139] Step 1.1: Calculate the power extraction ratio (PER) based on the voltage signal to be decomposed and multiple IMF components, and calculate the linear correlation between any two IMF components.

[0140] It should be understood that after each decomposition of the voltage signal to be decomposed using the VDM algorithm, relevant evaluation indicators (such as PER in this application, linear correlation between any two IMF components, etc.) of the multiple IMF components obtained by decomposition can be calculated, and the decomposition parameters (number of decompositions and penalty factor) can be dynamically adjusted according to the evaluation results. The voltage signal to be decomposed can then be re-decomposed according to the adjusted decomposition parameters so that the multiple IMF components obtained by re-decomposition are more accurate.

[0141] PER can be calculated using the following formula:

[0142]

[0143] In the above formula, The energy of the k-th IMF component. The total energy of the original signal x (i.e., the voltage signal to be decomposed) is given.

[0144] The linear correlation between any two IMF components can be calculated using the following formula:

[0145] ρ ij =|IMF i IMF j |

[0146] In the above formula, ρ ij This represents the correlation between the i-th IFM component and the j-th IFM component, where i ≠ j.

[0147] Step 1.2: If PER is less than the first preset value, the number of decompositions is increased by 1; if the maximum value of the linear correlation between any two IMF components is greater than the second preset value, the number of decompositions is decreased by 1.

[0148] In this step, if PER is less than the first preset value (e.g., 80%, the first preset threshold can be determined according to the actual application, and this application does not impose any restrictions on it), it can be determined that the decomposed modes are insufficient. The number of decomposed modes is dynamically increased by 1 to ensure that the main harmonic components of the voltage signal to be decomposed are fully extracted. If the maximum value of the linear correlation between any two IMF components is greater than the second preset value, it can be determined that the mode redundancy is high. The number of decomposed modes is dynamically decreased by 1 to prevent the voltage signal to be decomposed from being over-decomposed.

[0149] It should be understood that in other cases (i.e., PER is greater than or equal to the first preset value, and the maximum value of the linear correlation between any two IMF components is greater than or equal to the second preset threshold), the number of decompositions will remain unchanged.

[0150] Step 1.3: Based on the updated number of decompositions, the VMD algorithm is used to process the voltage signal to be decomposed to obtain multiple new IMF components corresponding to the voltage signal to be decomposed.

[0151] In this step, based on the adjusted number of decompositions and the current penalty factor, the voltage signal to be decomposed will be re-decomposed using the VMD algorithm to obtain multiple new IFM components, thereby improving the decomposition quality of the voltage signal to be decomposed.

[0152] The second aspect includes a method for dynamically adjusting the penalty factor and dynamically updating the multiple IMF components obtained from the decomposition based on the adjustment results:

[0153] Step 2.1: Calculate the short-time energy change rate ΔE of the voltage signal to be decomposed in real time.

[0154] Among them, the short-time energy change rate refers to the rate of change of the voltage signal energy within adjacent time windows. It is an indicator for measuring the degree of energy fluctuation of the voltage signal within adjacent time windows and can be obtained using algorithms commonly used in this field.

[0155] Step 2.2: If ΔE is not 0, use the formula α based on ΔE. new =α pre The penalty factor is updated using (1+0.1·tanh(ΔE / 0.2)), where α new Let α represent the updated penalty factor. pre This indicates the penalty factor before the update.

[0156] It should be understood that in practical applications, for microgrids requiring harmonic detection, the voltage signal of the microgrid will be acquired in real time. Because the voltage signal of the microgrid fluctuates during operation, the short-time energy change rate of the acquired voltage signal to be decomposed may change. By adjusting the penalty factor when the short-time energy change rate changes, the adaptability of the penalty factor to fluctuating voltage signals to be decomposed can be improved, further enhancing the decomposition accuracy of the voltage signal.

[0157] Step 2.3: Based on the updated penalty factor, the VMD algorithm is used to process the voltage signal to be decomposed to obtain multiple new IMF components corresponding to the voltage signal to be decomposed.

[0158] In this step, based on the adjusted penalty factor and the current number of decompositions, the voltage signal to be decomposed will be re-decomposed using the VMD algorithm to obtain multiple new IFM components, thereby improving the decomposition quality of the voltage signal to be decomposed.

[0159] In the two implementation methods provided in this embodiment, multiple evaluation indicators are calculated based on the multiple IFM components obtained from the decomposition and the voltage signal to be decomposed. Based on the corresponding adjustment rules, the number of decompositions and the penalty factor are dynamically adjusted from multiple dimensions to further improve the decomposition quality of the signal, thereby improving the accuracy of harmonic detection.

[0160] Furthermore, the method provided in this embodiment also includes:

[0161] Step 3.1: After each decomposition of the voltage signal to be decomposed using the VMD algorithm to obtain multiple IMF components, calculate the energy capture rate PER, modal redundancy (MRI), and signal-to-noise ratio (SNR) corresponding to the multiple IMF components.

[0162] The calculation method for PER has been explained above and will not be repeated here. The correlation and overlap between the multiple IMF components decomposed from the MRI response can be calculated using conventional algorithms in this field. SNR characterizes the relative intensity of useful components and noise components in a signal and can be defined as the ratio of signal power to noise power.

[0163] Step 3.2: Based on PER, MRI, and SNR, calculate the decomposition quality assessment index Q using the following formula:

[0164] Q = m1PER + m2(1-MRI) + m3SNR

[0165] Wherein, m1, m2 and m3 are the weighting coefficients of PER, MRI and SNR respectively, which can be flexibly adjusted according to the application scenario, and this application does not impose any restrictions on them; Q is used to indicate the quality of the IMF components obtained by decomposition.

[0166] Step 3.3: If Q exceeds the preset range, iteratively train the parameter prediction model.

[0167] The preset range refers to a pre-defined reasonable interval for Q. If the Q calculated from the multiple IMF components obtained from the decomposition is within the reasonable interval, it indicates that the decomposition of the voltage signal to be decomposed is appropriate. The preset range can be determined according to the actual application, and this application does not impose specific limitations.

[0168] It should be understood that if the Q values ​​corresponding to multiple IMF components obtained from the decomposition of the output of the parametric prediction model and the penalty factor decomposition exceed the preset range, it indicates that there is room for improvement in the prediction accuracy of the parametric prediction model. Similarly, if the Q values ​​corresponding to multiple IMF components obtained from the decomposition of the output of the parametric prediction model and the penalty factor decomposition exceed the preset range, there is still room for improvement in the prediction accuracy of the parametric prediction model. Therefore, after each decomposition of the voltage signal to be decomposed using the VMD algorithm to obtain multiple IMF components, if the calculated Q value exceeds the preset range, iterative training of the parametric prediction model should continue.

[0169] Optionally, if the number of times Q exceeds the preset range is greater than the preset number, then an incremental learning method is used to update the parameter prediction model.

[0170] It should be understood that if the number of times Q exceeds the preset range is greater than the preset number, it indicates that the number of decompositions and the penalty factor output by the parameter prediction model deviate significantly from the accurate value. In this case, an incremental learning method will be used to update the parameter prediction model based on newly acquired samples in order to ensure the accuracy of the model.

[0171] This embodiment improves the signal decomposition quality and thus the accuracy of harmonic detection by calculating multiple evaluation indicators based on the multiple IFM components and the voltage signal to be decomposed after each decomposition, and dynamically adjusting the number of decompositions and the penalty factor from multiple dimensions based on the corresponding adjustment rules. Furthermore, after each decomposition, the decomposition quality is evaluated according to the decomposition quality evaluation indicator Q, and corresponding training or reinforcement learning measures are taken for the model based on the evaluation results to ensure the accuracy of the model.

[0172] Figure 2 A schematic diagram illustrating the principle of the specific harmonic detection method provided in Embodiment 3 of this application is shown below. Figure 2As shown, in the method provided in this embodiment, the voltage signal to be decomposed is first obtained from the AC bus of the microgrid. The voltage signal to be decomposed is a three-phase voltage signal, including the a-phase voltage u. a Phase b voltage u b and the c-phase voltage u c The collected voltage signal to be decomposed in the abc coordinate system is processed and then transformed to the dq coordinate system to obtain the d component u of the voltage signal to be decomposed. d and q component u q Based on the d-component and q-component, feature extraction is performed on the voltage signal to be decomposed to obtain feature data. The feature data is then input into a pre-trained parameter prediction model to obtain the number of decompositions K and the penalty factor α. Based on the number of decompositions and the penalty factor, the variational mode decomposition (VMD) algorithm is used to process the voltage signal to be decomposed to obtain harmonic detection results (i.e., multiple intrinsic mode functions (IMF) components corresponding to the voltage signal to be decomposed). Based on the harmonic detection results, the number of decompositions K and the penalty factor α are dynamically adjusted. Based on the adjustment results, the VMD algorithm is used again to process the voltage signal to be decomposed to obtain new harmonic detection results. Simultaneously, based on the harmonic detection results, the quality of the decomposed IMF components is evaluated, and the parameter prediction model is adjusted accordingly based on the evaluation results.

[0173] The harmonic detection method provided in this application embodiment has a similar implementation principle and technical effect, and will not be described in detail here.

[0174] Figure 3 This is a schematic diagram of the structure of the harmonic detection device provided in Embodiment 4 of this application, as shown below. Figure 3 As shown, the harmonic detection device 20 provided in this embodiment includes:

[0175] The first acquisition unit 201 is used to acquire the voltage signal to be decomposed;

[0176] The first processing unit 202 is used to obtain feature data of the voltage signal to be decomposed based on the voltage signal to be decomposed.

[0177] The second processing unit 203 is used to input the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor; wherein, the parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm.

[0178] The third processing unit 204 is used to process the voltage signal to be decomposed using the variational mode decomposition (VMD) algorithm according to the number of decompositions and the penalty factor, so as to obtain multiple intrinsic mode function (IMF) components corresponding to the voltage signal to be decomposed.

[0179] The harmonic detection device 20 provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0180] Figure 4 This is a schematic diagram of the structure of the harmonic detection device provided in Embodiment 5 of this application, as shown below. Figure 4 As shown, based on the above embodiments, the harmonic detection device 20 provided in this embodiment further includes:

[0181] The second acquisition unit 205 is used to acquire a sample set, which includes: feature data corresponding to multiple different voltage signals, the number of calibrated decompositions, and a penalty factor; the multiple different voltage signals are collected when the microgrid is not connected to nonlinear devices and when it is connected to different nonlinear devices;

[0182] Training unit 206 is used to train the pre-constructed initial support vector regression model using the sample set to obtain the parameter prediction model.

[0183] The first update unit 207 is used to calculate the energy capture rate (PER) based on the voltage signal to be decomposed and multiple IMF components, and to calculate the linear correlation between any two IMF components; if the PER is less than a first preset value, the number of decompositions is incremented by 1; if the maximum value of the linear correlation between any two IMF components is greater than a second preset value, the number of decompositions is decremented by 1; based on the updated number of decompositions, the VMD algorithm is used to process the voltage signal to be decomposed to obtain new multiple IMF components corresponding to the voltage signal to be decomposed.

[0184] The second update unit 208 is used to calculate the short-time energy change rate ΔE of the voltage signal to be decomposed in real time.

[0185] If ΔE is not 0, then according to ΔE, the formula α is used. new =α pre The penalty factor is updated using (1 + 0.1 tanh(ΔE / 0.2)), where α new Let α represent the updated penalty factor. pre The penalty factor before the update is represented; based on the updated penalty factor, the VMD algorithm is used to process the voltage signal to be decomposed to obtain multiple new IMF components corresponding to the voltage signal to be decomposed.

[0186] The verification unit 209 is used to calculate the energy capture rate PER, modal redundancy MRI and signal-to-noise ratio SNR corresponding to the multiple IMF components after each decomposition of the voltage signal to be decomposed using the VMD algorithm.

[0187] Based on the PER, the MRI, and the SNR, the decomposition quality evaluation index Q is calculated using the formula Q = m1PER + m2(1-MRI) + m4SNR, where m1, m2, and m3 are the weighting coefficients of the PER, the MRI, and the SNR, respectively; and Q is used to indicate the quality of the IMF components obtained from the decomposition.

[0188] If Q exceeds a preset range, the parameter prediction model is iteratively trained.

[0189] The learning unit 210 is used to update the parameter prediction model by incremental learning if the number of times Q exceeds the preset range is greater than the preset number of times.

[0190] In one possible implementation, the kernel function in the initial support vector regression model of the training unit 206 is a hybrid kernel function obtained by combining the radial basis function kernel function and the sigmoid kernel function.

[0191] In one possible implementation, the second acquisition unit 205 is specifically used for:

[0192] Signals were acquired from the AC bus of the microgrid under two conditions: with no nonlinear devices connected and with different nonlinear devices connected. Multiple voltage signals were obtained.

[0193] For each voltage signal, characteristic data of the voltage signal are obtained based on the voltage signal;

[0194] Based on the grid search algorithm, the number of decompositions and the penalty factor corresponding to the voltage signal are determined;

[0195] The sample set is constructed based on the feature data of each voltage signal, the number of decompositions corresponding to each voltage signal, and the penalty factor.

[0196] In one possible implementation, the feature data includes at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

[0197] In one possible implementation, the training unit 206 selects a loss function during model training. for:

[0198]

[0199] Where N represents the number of training samples, w i Let y represent the weight of the i-th sample. i This represents the true label of the i-th sample. Let represent the predicted label of the i-th sample, and ε represent the threshold.

[0200] The harmonic detection device 20 provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0201] Figure 5 A schematic diagram of the structure of the computer device provided in this application. Figure 5 As shown, the computer device 30 provided in this embodiment includes at least one processor 301 and a memory 302. Optionally, the computer device 30 further includes a communication component 303. The processor 301, memory 302, and communication component 303 are connected via a bus 304.

[0202] In a specific implementation, at least one processor 301 executes computer execution instructions stored in memory 302, causing at least one processor 301 to perform the above-described method.

[0203] The specific implementation process of processor 301 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0204] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0205] The memory may include read-only memory and random access memory. The memory may be volatile or non-volatile, or may include both. Non-volatile memory may include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which serves as an external cache. Many forms of RAM are available by way of example, but not limitation. Examples include Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0206] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0207] This application also provides a computer program product, including a computer program that, when executed, implements the above-described method.

[0208] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed, implement the above-described method.

[0209] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as SRAM, EEPROM, EPROM, PROM, ROM, magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0210] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside within an ASIC. Alternatively, the processor and the readable storage medium can exist as discrete components in a device.

[0211] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0212] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0213] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0214] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0215] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0216] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A harmonic detection method, characterized in that, include: Acquire the voltage signal to be decomposed; Based on the voltage signal to be decomposed, obtain the feature data of the voltage signal to be decomposed; The feature data is input into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor; wherein, the parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a machine learning algorithm. Based on the number of decompositions and the penalty factor, the variational mode decomposition (VMD) algorithm is used to process the voltage signal to be decomposed, thereby obtaining multiple intrinsic mode function (IMF) components corresponding to the voltage signal to be decomposed.

2. The method according to claim 1, characterized in that, Before inputting the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor, the method further includes: A sample set is obtained, which includes: feature data corresponding to multiple different voltage signals, the number of calibrated decompositions, and penalty factors; the multiple different voltage signals are collected when the microgrid is not connected to nonlinear devices and when it is connected to different nonlinear devices; The parameter prediction model is obtained by training the pre-constructed initial support vector regression model using the sample set.

3. The method according to claim 2, characterized in that, The kernel function in the initial support vector regression model is a hybrid kernel function obtained by combining the radial basis function and the sigmoid kernel function.

4. The method according to claim 2, characterized in that, The acquisition of the sample set includes: Signals were acquired from the AC bus of the microgrid under two conditions: with no nonlinear devices connected and with different nonlinear devices connected. Multiple voltage signals were obtained. For each voltage signal, characteristic data of the voltage signal are obtained based on the voltage signal; Based on the grid search algorithm, the number of decompositions and the penalty factor corresponding to the voltage signal are determined; The sample set is constructed based on the feature data of each voltage signal, the number of decompositions corresponding to each voltage signal, and the penalty factor.

5. The method according to any one of claims 1 to 4, characterized in that, The feature data includes at least one of zero-crossing rate, spectral skewness, fuzzy entropy, and wavelet packet energy entropy.

6. The method according to claim 2, characterized in that, The method further includes: The loss function selected during model training for: Where N represents the number of training samples, w i Let y represent the weight of the i-th sample. i This represents the true label of the i-th sample. Let represent the predicted label of the i-th sample, and ε represent the threshold.

7. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Based on the voltage signal to be decomposed and multiple IMF components, the energy capture rate PER is calculated, and the linear correlation between any two IMF components is calculated. If the PER is less than the first preset value, then the number of decompositions is incremented by 1; If the maximum value of the linear correlation between any two IMF components is greater than the second preset value, then the number of decompositions is reduced by 1. Based on the updated number of decompositions, the VMD algorithm is used to process the voltage signal to be decomposed, and new IMF components corresponding to the voltage signal to be decomposed are obtained.

8. The method according to any one of claims 1-4, characterized in that, The method further includes: The short-time energy change rate ΔE of the voltage signal to be decomposed is calculated in real time. If ΔE is not 0, then according to ΔE, the formula α is used. new =α pre The penalty factor is updated using (1 + 0.1 tanh(ΔE / 0.2)), where α new Let α represent the updated penalty factor. pre Indicates the penalty factor before the update; Based on the updated penalty factor, the VMD algorithm is used to process the voltage signal to be decomposed, and new IMF components corresponding to the voltage signal to be decomposed are obtained.

9. The method according to any one of claims 1-4, characterized in that, The method further includes: After each decomposition of the voltage signal to be decomposed using the VMD algorithm to obtain multiple IMF components, the energy capture rate PER, modal redundancy MRI, and signal-to-noise ratio SNR corresponding to the multiple IMF components are calculated. Based on the PER, the MRI, and the SNR, the decomposition quality evaluation index Q is calculated using the formula Q = m1PER + m2(1-MRI) + m3SNR, where m1, m2, and m3 are the weighting coefficients of the PER, the MRI, and the SNR, respectively; Q is used to indicate the quality of the IMF components obtained from the decomposition. If Q exceeds a preset range, the parameter prediction model is iteratively trained.

10. The method according to claim 9, characterized in that, The method further includes: If the number of times Q exceeds the preset range is greater than the preset number, then incremental learning is used to update the parameter prediction model.

11. A harmonic detection device, characterized in that, include: The first acquisition unit is used to acquire the voltage signal to be decomposed; The first processing unit is used to obtain the feature data of the voltage signal to be decomposed based on the voltage signal to be decomposed. The second processing unit is used to input the feature data into a pre-trained parameter prediction model to obtain the number of decompositions and the penalty factor; wherein, the parameter prediction model is a model that predicts decomposition parameters based on voltage signal features and is pre-trained based on a support vector regression model. The third processing unit is used to process the voltage signal to be decomposed using the variational mode decomposition (VMD) algorithm according to the number of decompositions and the penalty factor, so as to obtain multiple intrinsic mode function (IMF) components corresponding to the voltage signal to be decomposed.

12. A computer device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-10.

Images