Power transient signal detection method and system based on optimized decision network

Abstract

The present application provides a method and system for detecting power transient signals based on an optimized decision network. The method for detecting power transient signals based on an optimized decision network includes: collecting transient signals of a power system; denoising the transient signals by wavelet transform based on an absolute median deviation threshold; performing state detection on the denoised transient signals based on the absolute median deviation, and obtaining a first state detection result; performing state detection on the denoised transient signals based on an artificial neural network with an attention mechanism added, and obtaining a second state detection result; performing an OR gate operation on the first state detection result and the second state detection result, and outputting the detection result, so as to solve the technical problems that the existing methods generally have a large amount of calculation and cannot detect quickly.

Description

Technical Field

[0001] The present application relates to the technical field of state detection of transient signals in power systems, and in particular to a method and system for detecting transient signals in power systems based on an optimized decision network. Background Art

[0002] The widespread use of power electronic equipment and the increase in nonlinear loads have made transient power quality issues increasingly prominent, and have attracted great attention from power departments and users. Improving power quality has a positive effect on the safe operation of the power grid, the normal operation of industrial production, and energy conservation and emission reduction.

[0003] Detection of transient signals in power systems mainly refers to obtaining the start and end times of disturbances, which is used as the basic content of power quality analysis. For state detection of transient signals, domestic and foreign scholars have proposed many methods, such as short-time Fourier transform (STFT), S transform, atomic decomposition, wavelet transform, etc.

[0004] Among them, the detection method based on wavelet transform mainly uses the modulus maximum of wavelet transform to distinguish steady-state disturbance from transient disturbance, and further realizes the state detection of transient signal. The S-transform Gaussian window function has a fixed shape and the time-frequency resolution change trend remains unchanged. However, it is difficult to select the adjustment parameters of the improved S-transform Gaussian window function, and the S-transform has a large amount of calculation, which is not suitable for use in engineering practice.

[0005] These methods are based on a single disturbance to detect the state of transient signals. However, in practice, the grid voltage often exists in the form of a combination of multiple disturbances such as noise, harmonics, and electromagnetic transients. The overlap of various characteristic quantities brings great difficulties to the detection, making these methods based on a single disturbance ineffective.

[0006] With the study of transient signals, scholars have proposed state detection methods for composite transient signals. For example, the composite disturbance signal is first subjected to morphological filtering and wavelet multi-resolution decomposition, and then the Teager energy operator and Prony algorithm are used to extract the amplitude, frequency, and attenuation factor of each frequency band, or the energy distribution of different frequency bands, wavelet entropy, and root mean square value are used as feature vectors for disturbance judgment, and then combined with neural networks for classification to achieve accurate detection and identification of each disturbance under composite disturbance. However, such methods generally have the problem of large computational complexity and inability to detect quickly. Summary of the invention

[0007] The present application provides a method and system for detecting power transient signals based on an optimized decision network, so as to solve the technical problems that the existing methods generally have large computational complexity and cannot detect quickly.

[0008] The present application provides a method for detecting power transient signals based on an optimized decision network, comprising:

[0009] Collect transient signals of power system;

[0010] De-noising transient signals using wavelet transform based on absolute median deviation threshold;

[0011] Performing state detection on the denoised transient signal based on the absolute median deviation, and obtaining a first state detection result;

[0012] Based on the artificial neural network with attention mechanism, the state of the denoised transient signal is detected, and the second state detection result is obtained;

[0013] Perform an OR gate operation on the first state detection result and the second state detection result, and output the detection result.

[0014] Optionally, the step of performing denoising on the transient signal based on wavelet transform of absolute median deviation threshold comprises the following steps:

[0015] Decompose the transient signal by wavelet, and obtain the real signal and the noise signal, the wavelet coefficient of the real signal is greater than the wavelet coefficient of the noise signal;

[0016] Construct a threshold based on the absolute median deviation;

[0017] Apply soft threshold function for wavelet denoising, the soft threshold function is:

[0018]

[0019] Among them, sign(·) represents the sign function, represents the updated wavelet coefficient, λ represents the threshold, D j (k) represents the high frequency coefficient of the jth layer, N represents the signal length;

[0020] Reconstruct the denoised transient signal.

[0021] Optionally, in the step of performing wavelet decomposition on the transient signal, the transient signal is decomposed on the basis of selecting Symlets wavelet as the basis function, and the high-frequency coefficients of the real signal and the low-frequency coefficients of the noise signal are obtained.

[0022]

[0023] Among them, the input transient signal x(n) with length N, C j (k) represents the low-frequency coefficient of the jth layer, D j(k) represents the high-frequency coefficient of the jth layer, h(n) and g(n) represent the low-pass and high-pass filter coefficients respectively, N represents the length of the transient signal, n represents the discrete points in the transient signal,

[0024] Optionally, in the step of constructing a threshold based on the absolute median deviation, the absolute median deviation is:

[0025] MAD(D j (k))=median{|D j (k)-median(D j (k))|}

[0026] Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0027] The improved absolute median deviation formula is:

[0028]

[0029] Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0030] The formula for the threshold is:

[0031] λ=μ1MAD(D j (k))+μ2MAD_N(D j (k))

[0032] Where λ represents the threshold, μ1 and μ2 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0033] Optionally, in the step of performing state detection on the denoised transient signal based on the absolute median deviation and obtaining a first state detection result,

[0034] A transient signal sequence W containing the latest m samples of X within an observation interval:

[0035] W=[X(tm·Δt),...,X(t-Δt)X(t)]

[0036] Where Δt is a sampling interval, W is the transient signal sequence, X is the observation interval, and m is the number of samples;

[0037] The absolute median deviation and improved absolute median deviation of the transient signal sequence W are:

[0038] MAD(W)=median{|W-median(W)|}

[0039]

[0040] Where median(·) represents the median operation, W is the transient signal sequence, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0041] The first threshold is constructed according to the absolute median deviation and the improved absolute median deviation, and the first threshold is:

[0042] λ1=μ3MAD(W)+μ4MAD_N(W)

[0043] Wherein, λ1 represents the first threshold, μ3 and μ4 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, W is the transient signal sequence,

[0044] Calculating the derivative of the difference between the transient signal sequence and the first threshold value;

[0045] It is determined whether the derivative is greater than a preset value. If so, the first state detection result is an abnormal state. If not, the first state detection result is a normal state.

[0046] Optionally, the step of performing state detection on the denoised transient signal based on the artificial neural network with the attention mechanism added and obtaining the second state detection result includes the following steps:

[0047] According to the denoised transient signal, a feature vector is obtained, which includes the mean value, kurtosis, shape factor, absolute median deviation, and improved absolute median deviation of the transient signal. The feature matrix is ​​as follows:

[0048]

[0049] Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals;

[0050] 80% of the entire data set is randomly selected for artificial neural network training, and the rest is used for artificial neural network testing; the formula for data setting is:

[0051]

[0052] Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals;

[0053] Add an attention mechanism to the artificial neural network and input the feature vector into the attention mechanism;

[0054] The output value of the artificial neural network is compared with the preset output value and the error is calculated using the cross entropy loss function;

[0055] The error is propagated back to the network and the weights are set according to the error, and the second state detection result is obtained.

[0056] Optionally, the artificial neural network includes more than two processing units. For a certain processing unit, the information from other processing units i is defined as X i , the interaction strength between them and the processing unit, i.e. the connection weight, is W i , i=0,1,…,n-1, the internal threshold of the processing unit is θ;

[0057] The input of this processing unit is: The output is Among them, X i is the input of the i-th element, W i is the interconnection weight between the ith processing unit and this processing unit, f is the activation function, which determines the output of the processing unit, and θ represents the threshold of the hidden layer neural node.

[0058] Correspondingly, the present application also provides an electronic device, which includes a memory and a processor, the memory being used to store executable program code; the processor being connected to the memory, and running a computer program corresponding to the executable program code by reading the executable program code, so as to execute any of the steps in the above-mentioned method for detecting power transient signals based on an optimized decision network.

[0059] Correspondingly, the present application also provides a power transient signal detection system based on an optimized decision network, which includes the electronic device described above.

[0060] The present application provides a method and system for detecting power transient signals based on an optimized decision network. The transient signals are first denoised using wavelet denoising, and then an optimized decision network is constructed using absolute median deviation and an artificial neural network to detect the state of the transient signals. In addition, by adding an attention mechanism module to the artificial neural network, the optimized decision network is more focused on the characteristic information of abnormal transient signals, which improves the accuracy of transient signal recognition, so that the method can accurately and quickly detect the state of transient signals in the presence of noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0062] Figure 1 It is a flow chart of a method for detecting power transient signals based on an optimized decision network provided by the present application;

[0063] Figure 2 It is a flow chart of step S200 in the power transient signal detection method based on the optimization decision network provided by the present application;

[0064] Figure 3 It is a schematic diagram of mixed disturbance samples in the power transient signal detection method based on optimized decision network provided in this application. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application. In addition, it should be understood that the specific implementation methods described herein are only used to illustrate and explain the present application and are not used to limit the present application. In the present application, unless otherwise stated, the directional words used, such as "up", "down", "left", and "right", generally refer to the up, down, left, and right of the device in actual use or working state, specifically the drawing direction in the accompanying drawings.

[0066] The present application provides a method and system for detecting power transient signals based on an optimized decision network, which are described in detail below. It should be noted that the description order of the following embodiments is not intended to limit the preferred order of the embodiments of the present application. In the following embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0067] See also Figure 1The present application provides a method for detecting power transient signals based on an optimized decision network. First, wavelet denoising is used to preprocess the transient signals. At the same time, absolute median deviation and artificial neural network are used as optimized decision networks to detect the state of transient signals. In the optimized decision network, the absolute median deviation first detects the state of transient signals and extracts features from transient signals. Then, an artificial neural network with an attention mechanism is used to perform feature analysis and detection. Finally, an OR operation is performed on the two state detections to realize state detection of transient signals in power systems, so that the method has better robustness and performance.

[0068] See also Figure 1 and Figure 2 , a method for detecting power transient signals based on an optimized decision network, specifically comprising the following steps:

[0069] S100, collecting transient signals of the power system;

[0070] S200, performing denoising processing on the transient signal based on wavelet transform of absolute median deviation threshold;

[0071] After wavelet transforming the transient signal containing noise, the obtained wavelet coefficients contain important information about the real signal and the noise signal. Among them, the wavelet coefficients of the real signal are larger after wavelet decomposition, and the wavelet coefficients of the noise signal are smaller after wavelet decomposition. The wavelet coefficients of the real signal are much larger than those of the noise signal. Therefore, according to this feature, a suitable threshold can be selected to make the wavelet coefficients of the noise signal 0, eliminate the influence of noise, and make the reconstructed signal free of noise, thereby achieving the purpose of denoising.

[0072] Step S200 specifically includes the following steps:

[0073] S210, performing wavelet decomposition on the transient signal to obtain a real signal and a noise signal, wherein the wavelet coefficient of the real signal is greater than the wavelet coefficient of the noise signal;

[0074] The transient signal is decomposed based on the selection of Symlets wavelet as the basis function, and the high-frequency coefficients of the real signal and the low-frequency coefficients of the noise signal are obtained.

[0075]

[0076] Among them, the input transient signal x(n) with length N, C j (k) represents the low-frequency coefficient of the jth layer, D j (k) represents the high-frequency coefficient of the jth layer, h(n) and g(n) represent the low-pass and high-pass filter coefficients respectively, N represents the length of the transient signal, n represents the discrete points in the transient signal,

[0077] S220, constructing a threshold based on the absolute median deviation;

[0078] The voltage disturbance signal is decomposed based on the selection of Symlets wavelet as the basis function. For the decomposed transient signal, the wavelet coefficient of the noise is smaller than the wavelet coefficient of the signal. This application selects a suitable threshold to make the wavelet coefficient of the noise signal 0. The threshold is selected as follows:

[0079] In the step of constructing a threshold based on the absolute median deviation, the absolute median deviation is:

[0080] MAD(D j (k))=median{|D j (k)-median(D j (k))|}

[0081] Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0082] The improved absolute median deviation formula is:

[0083]

[0084] Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0085] The improved absolute median deviation can better adapt to the noise intensity in transient signals and locate outliers in wavelet coefficients, that is, the values ​​with the most sudden changes in the transient process.

[0086] The formula for the threshold is:

[0087] λ=μ1MAD(D j (k))+μ2MAD_N(D j (k))

[0088] Where λ represents the threshold, μ1 and μ2 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0089] S230, applying a soft threshold function to perform wavelet denoising, the soft threshold function is:

[0090]

[0091] Among them, sign(·) represents the sign function, represents the updated wavelet coefficient, λ represents the threshold, D j (k) represents the high frequency coefficient of the jth layer, N represents the signal length;

[0092] S240, reconstructing the denoised transient signal;

[0093] In wavelet transform denoising, a soft threshold function is selected, in which the absolute median deviation and the improved absolute median deviation are used to adaptively determine the threshold. The adaptively determined threshold makes the wavelet coefficient of the noise signal equal to 0, thereby completing the reconstruction of the transient signal and finally successfully filtering out the noise in the transient signal.

[0094] S300, performing state detection on the denoised transient signal based on the absolute median deviation, and obtaining a first state detection result;

[0095] The median absolute deviation method is a robust statistical method that can locate the most abrupt values ​​in transient processes in a dataset or signal sequence.

[0096] According to the mixed disturbance signal model, a series of mixed disturbance samples can be generated in batches with the help of Matlab script program. The typical waveform is as follows Figure 3 As shown in Figure 2. When disturbance occurs, the signal has an obvious mutation process. In view of this feature, we use the absolute median deviation and improved absolute median deviation methods to detect the state of transient signals.

[0097] Assume that a transient signal sequence W is given, which contains the latest m samples of X in an observation interval:

[0098] W=[X(tm·Δt),...,X(t-Δt)X(t)]

[0099] Where Δt is a sampling interval, W is the transient signal sequence, X is the observation interval, and m is the number of samples;

[0100] The absolute median deviation and improved absolute median deviation of the transient signal sequence W are:

[0101] MAD(W)=median{|W-median(W)|}

[0102]

[0103] Where median(·) represents the median operation, W is the transient signal sequence, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal,

[0104] The first threshold is constructed according to the absolute median deviation and the improved absolute median deviation, and the first threshold is:

[0105] λ1=μ3MAD(W)+μ4MAD_N(W)

[0106] Wherein, λ1 represents the first threshold, μ3 and μ4 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, W is the transient signal sequence,

[0107] Calculating the derivative of the difference between the transient signal sequence and the first threshold value;

[0108] It is determined whether the derivative is greater than a preset value. If so, the first state detection result is an abnormal state. If not, the first state detection result is a normal state.

[0109] In view of the fact that transient signals have obvious mutation processes when disturbances occur, the present application adds the absolute median deviation and the improved absolute median deviation and sets their respective weights to form a criterion for the disturbance of transient signals in the power system, accurately detects abnormal values ​​in transient signals in the power system, and thus realizes the state detection of transient signals.

[0110] S400, performing state detection on the denoised transient signal based on an artificial neural network with an attention mechanism added, and obtaining a second state detection result;

[0111] Artificial neural network is an artificial network composed of a large number of processing units that are widely interconnected. For a certain processing unit (neuron), suppose the information from other processing units (neurons) i is X i , the interaction strength between them and the processing unit, i.e. the connection weight, is W i , i=0,1,…,n-1, the internal threshold of the processing unit is θ. The input of this processing unit (neuron) is: The output is Among them, X i is the input of the i-th element, W i is the interconnection weight between the ith processing unit and the current processing unit. f is called the activation function, which determines the output of the node (neuron). θ represents the threshold of the hidden layer neural node. The artificial neural network calculates the output result and corrects the weight according to the error between the actual output and the expected output. When the artificial neural network judges the state of the transient signal, it only inputs the transient signal but not the expected output result.

[0112] The attention mechanism neural network uses the features extracted from transient signals to classify states. The training process of the attention mechanism neural network is as follows:

[0113] S410. Obtain a feature vector according to the denoised transient signal. The feature vector includes the mean value, kurtosis, shape factor, absolute median deviation, and improved absolute median deviation of the transient signal. The feature matrix is ​​as follows:

[0114]

[0115] Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals;

[0116] Using the feature vectors of transient signals such as mean, kurtosis, shape factor, absolute median deviation, and improved absolute median deviation can ensure that the artificial neural network can discard repeated redundant information in the input data.

[0117] S420, randomly select 80% from the entire data set for artificial neural network training, and the rest for artificial neural network testing; the formula for data setting is:

[0118]

[0119] Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals;

[0120] S430, adding an attention mechanism to the artificial neural network, and inputting the feature vector into the attention mechanism;

[0121] S440, comparing the output value of the artificial neural network with the preset output value, and calculating the error using a cross entropy loss function;

[0122] S450, propagating the error back to the network and setting the weight according to the error, and obtaining a second state detection result;

[0123] By adding the attention mechanism, the artificial neural network can focus more on the feature changes when the transient signal is disturbed during the process of updating the weights, making it easier to identify the state changes of the transient signal.

[0124] This application proposes a method of applying the attention mechanism to an artificial neural network, using the rate of change of each feature when a disturbance occurs as a weighted condition, successfully extracting important features and increasing the recognition rate of state detection.

[0125] S500, performing an OR gate operation on the first state detection result and the second state detection result, and outputting the detection result;

[0126] The present application increases the accuracy of transient signal detection and makes the algorithm more robust by using the improved absolute median deviation and the detection result of the artificial neural network for OR gate operation.

[0127] The present application also provides a power transient signal detection system based on an optimized decision network, which includes the electronic device described above. The electronic device includes a memory and a processor, the memory is used to store executable program codes; the processor is connected to the memory, and runs a computer program corresponding to the executable program code by reading the executable program code to execute the steps in the power transient signal detection method based on the optimized decision network described above.

[0128] The above is a detailed introduction to the power transient signal detection method and system based on the optimized decision network provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for detecting power transient signals based on an optimized decision network, characterized in that: include: Collect transient signals of power system; De-noising transient signals using wavelet transform based on absolute median deviation threshold; Performing state detection on the denoised transient signal based on the absolute median deviation, and obtaining a first state detection result; Based on the artificial neural network with attention mechanism, the state of the denoised transient signal is detected, and the second state detection result is obtained; Perform an OR gate operation on the first state detection result and the second state detection result, and output the detection result. Wherein, in the step of performing state detection on the denoised transient signal based on the absolute median deviation and obtaining a first state detection result, A transient signal sequence W containing the latest m samples of X within an observation interval: W=[X(tm·Δt),…,X(t-Δt),X(t)] Where Δt is a sampling interval, W is the transient signal sequence, X is the observation interval, and m is the number of samples; The absolute median deviation and improved absolute median deviation of the transient signal sequence W are: MAD(W)=median{W-median(W)} Where median(·) represents the median operation, W is the transient signal sequence, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal, The first threshold is constructed according to the absolute median deviation and the improved absolute median deviation, and the first threshold is: λ1=μ3MAD(W)+μ4MAD - N(W) Wherein, λ1 represents the first threshold, μ3 and μ4 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, W is the transient signal sequence, Calculating the derivative of the difference between the transient signal sequence and the first threshold value; Determine whether the derivative is greater than a preset value. If so, the first state detection result is an abnormal state. If not, the first state detection result is a normal state. The artificial neural network based on the attention mechanism performs state detection on the denoised transient signal and obtains a second state detection result, including the following steps: According to the denoised transient signal, a feature vector is obtained, which includes the mean value, kurtosis, shape factor, absolute median deviation, and improved absolute median deviation of the transient signal. The feature matrix is ​​as follows: Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals; 80% of the entire data set is randomly selected for artificial neural network training, and the rest is used for artificial neural network testing; the formula for data setting is: Among them, F1, F2, F3, F4, and F5 represent feature vectors extracted from transient signals; Add an attention mechanism to the artificial neural network and input the feature vector into the attention mechanism; The output value of the artificial neural network is compared with the preset output value and the error is calculated using the cross entropy loss function; The error is propagated back to the network and the weights are set according to the error, and the second state detection result is obtained.

2. The method for detecting power transient signals based on an optimized decision network according to claim 1, characterized in that: The step of performing denoising on the transient signal based on wavelet transform of absolute median deviation threshold comprises the following steps: Decompose the transient signal by wavelet, and obtain the real signal and the noise signal, the wavelet coefficient of the real signal is greater than the wavelet coefficient of the noise signal; Construct a threshold based on the absolute median deviation; Apply soft threshold function for wavelet denoising, the soft threshold function is: Among them, sign(·) represents the sign function, represents the updated wavelet coefficient, λ represents the threshold, D j (k) represents the high frequency coefficient of the jth layer, N represents the signal length; Reconstruct the denoised transient signal.

3. The power transient signal detection method based on the optimization decision network according to claim 2 is characterized in that: In the step of performing wavelet decomposition on the transient signal, the transient signal is decomposed on the basis of selecting the Symlets wavelet as the basis function, and the high-frequency coefficients of the real signal and the low-frequency coefficients of the noise signal are obtained. Among them, the input transient signal x(n) with length N, C j (k) represents the low-frequency coefficient of the jth layer, D j (k) represents the high-frequency coefficient of the jth layer, h(n) and g(n) represent the low-pass and high-pass filter coefficients respectively, N represents the length of the transient signal, n represents the discrete points in the transient signal, 4. The method for detecting power transient signals based on an optimized decision network according to claim 2, characterized in that: In the step of constructing a threshold based on the absolute median deviation, the absolute median deviation is: MAD(D j (k))=median{D j (k)-median(D j (k))} Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal, The improved absolute median deviation formula is: Among them, median(·) represents the median operation, D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal, The formula for the threshold is: λ=μ1MAD(D j (k))+μ2MAD - N(D j (k)) Where λ represents the threshold, μ1 and μ2 represent the parameters for regulating the absolute median deviation and the improved absolute median deviation ratio, and D j (k) represents the high frequency coefficient of the jth layer, N represents the length of the transient signal, 5. The method for detecting power transient signals based on an optimized decision network according to claim 1, characterized in that: The artificial neural network consists of two or more processing units. For a certain processing unit, the information from other processing units i is defined as X i , the interaction strength between them and the processing unit, i.e. the connection weight, is W i , i=0,1,…,n-1, the internal threshold of the processing unit is θ; The input of this processing unit is: The output is Among them, X i is the input of the i-th element, W i is the interconnection weight between the ith processing unit and this processing unit, f is the activation function, which determines the output of the processing unit, and θ represents the threshold of the hidden layer neural node.

6. An electronic device, characterized in that: include: A memory for storing executable program codes; as well as A processor is connected to the memory, and runs a computer program corresponding to the executable program code by reading the executable program code to execute the steps in the power transient signal detection method based on the optimized decision network as claimed in any one of claims 1 to 5.

7. A power transient signal detection system based on an optimized decision network, characterized in that: Comprising the electronic device as claimed in claim 6.

Images