CN116559728A - Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron - Google Patents
Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron Download PDFInfo
- Publication number
- CN116559728A CN116559728A CN202310538831.3A CN202310538831A CN116559728A CN 116559728 A CN116559728 A CN 116559728A CN 202310538831 A CN202310538831 A CN 202310538831A CN 116559728 A CN116559728 A CN 116559728A
- Authority
- CN
- China
- Prior art keywords
- wavelet
- formula
- denoising
- fault
- layer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000005540 biological transmission Effects 0.000 title claims abstract description 33
- 210000002569 neuron Anatomy 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 25
- 239000011159 matrix material Substances 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 230000000694 effects Effects 0.000 claims abstract description 11
- 238000005457 optimization Methods 0.000 claims abstract description 6
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 42
- 230000008569 process Effects 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000013139 quantization Methods 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 13
- 238000010801 machine learning Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 5
- 238000007637 random forest analysis Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 2
- 101100499229 Mus musculus Dhrsx gene Proteins 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 230000002035 prolonged effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 229920003020 cross-linked polyethylene Polymers 0.000 description 1
- 239000004703 cross-linked polyethylene Substances 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/50—Testing of electric apparatus, lines, cables or components for short-circuits, continuity, leakage current or incorrect line connections
- G01R31/58—Testing of lines, cables or conductors
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
- Y04S10/52—Outage or fault management, e.g. fault detection or location
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention discloses a power transmission line fault diagnosis method based on a wavelet transformation-multilayer perceptron, which comprises the following steps: 1) Collecting an original three-phase current signal of a power transmission system, and numbering fault types; then, carrying out noise adding treatment on the original three-phase current signals, and dividing the samples into training sets and test sets; 2) Acquiring wavelet coefficients of different frequency bands through wavelet decomposition, determining a wavelet denoising threshold and a denoising mode to perform threshold denoising, and selecting a wavelet basis function with the best denoising effect to perform wavelet decomposition; 3) Extracting wavelet coefficient energy values, constructing a feature matrix of fault energy, preprocessing the feature matrix in different modes, and selecting a data processing mode with highest diagnosis accuracy; 4) And (3) adjusting the structure of the network full-connection layer and the number of neurons, and adding Dropout network optimization to obtain an MLP fault diagnosis model. The method reduces the influence of noise and obviously improves the training speed and the diagnosis accuracy.
Description
Technical Field
The invention belongs to the technical field of fault diagnosis of power transmission lines, and relates to a power transmission line fault diagnosis method based on a wavelet transformation-multilayer perceptron.
Background
Along with the continuous development of the power system, the method realizes the rapid diagnosis and early warning of faults in the power transmission line, and has important significance in ensuring the reliable operation of the power system. However, the existing fault diagnosis technology has the problems of complex network model, low diagnosis speed and low diagnosis accuracy. Aiming at the problem, if the early fault characteristics in the monitoring data can be found by excavating the real-time monitoring data before the permanent fault occurs, real-time diagnosis and early warning are carried out, an adjustable overhaul plan is provided for operation and maintenance personnel, and the safe and stable operation of the power system is ensured.
Recently, deep learning is increasingly widely applied to fault diagnosis of a power system, and the fault on-line diagnosis can be realized by using classical multi-layer perceptrons (Multilayer Perceptron, MLP) in the deep learning. The high-dimensional characteristics in the fault data are extracted through multiple full-connection nonlinear operations, so that accurate diagnosis of complex faults is realized, however, original data are used as diagnosis network input, redundant information is more, model training time is prolonged, and the accuracy of a diagnosis model is influenced by noise contained in real-time monitoring data.
Disclosure of Invention
The invention aims to provide a power transmission line fault diagnosis method based on a wavelet transformation-multilayer perceptron, which solves the problems that in the prior art, redundant information is more, model training time is prolonged, noise in monitoring data is more, and power transmission line fault diagnosis accuracy cannot meet technical requirements.
The technical scheme adopted by the invention is that the power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron is implemented according to the following steps:
step 1, acquiring an original three-phase current signal of a power transmission system, and numbering fault types to determine fault type numbers corresponding to the original three-phase current signal; then, carrying out noise adding processing on the original three-phase current signal to obtain a fault sample containing noise, and dividing the sample into a training set and a testing set;
step 2, acquiring wavelet coefficients of different frequency bands through wavelet decomposition, determining a wavelet denoising threshold and a denoising mode to perform threshold denoising, and selecting a wavelet basis function with the best denoising effect to perform wavelet decomposition by verifying denoising effects of different wavelet basis functions;
step 3, extracting wavelet coefficient energy values, constructing a feature matrix of fault energy, preprocessing the feature matrix in different modes, and selecting a data processing mode with highest diagnosis accuracy;
and 4, establishing an MLP fault diagnosis model, adjusting the network full-connection layer structure and the number of neurons, adding Dropout network optimization, and taking the one with the highest fault diagnosis accuracy as the MLP fault diagnosis model.
The beneficial effects of the invention are that the invention comprises the following aspects:
1) The method combines the advantages of the two methods, noise suppression is carried out through wavelet change, the influence of noise on a diagnosis model is reduced, the extracted energy characteristic matrix can be used as input of a diagnosis network, and the energy values of all coefficients can be further analyzed.
2) The multi-layer perceptron model is built for fault diagnosis, the characteristic energy matrix is used as network input, standard normal distribution is selected for processing, the model calculation load can be reduced, the model training speed and the diagnosis accuracy can be remarkably improved, and a reference is provided for fault diagnosis of the power transmission line.
3) The fault diagnosis accuracy of the power transmission line provided by the invention is superior to that of the traditional machine learning method, and can realize more efficient fault diagnosis.
Drawings
FIG. 1 is a schematic diagram of an embodiment architecture of a multi-layer perceptron employed by the present invention;
FIG. 2a is a Dropout network optimization criteria full connectivity layer network; FIG. 2b is a full connectivity layer network using Dropout technology;
FIG. 3 is an overall flow diagram of the method of the present invention;
FIG. 4a is a graph of the multiple cycle early fault phase A current wavelet coefficient energy values; FIG. 4b is a half cycle early fault phase A current wavelet coefficient energy value; FIG. 4c is a constant impedance ground fault phase A current wavelet coefficient energy value;
FIG. 5a is a schematic diagram of an electrical diagram model multiple cycle early failure; FIG. 5b is a schematic diagram of half cycle early failure; FIG. 5c is a constant impedance ground schematic;
FIG. 6 is a schematic diagram of a fault raw energy value for an embodiment of the present invention;
FIG. 7 is a schematic diagram of microwave energy pretreatment according to an embodiment of the present invention;
FIG. 8 is a training accuracy diagram of an embodiment of the present invention;
FIG. 9a is a raw current signal identification result; FIG. 9b is a normalized energy signature recognition result of an embodiment of the present invention;
FIG. 10a is a graph showing the diagnostic results of the method of the present invention, and FIG. 10b is a graph showing the diagnostic results of the method of the present invention compared to the decision tree; FIG. 10c is a graph comparing the diagnostic results of the method of the present invention with the K nearest neighbor.
Detailed Description
The invention will be described in detail below with reference to the drawings and the detailed description.
As a classical deep learning network, the multi-layer perceptron has high response speed and low calculation requirement, and the most suitable network model is selected through model structure and parameter adjustment, so that the response speed and accuracy of line fault diagnosis are obviously improved. In order to reduce the influence of noise on the fault diagnosis accuracy of the multi-layer perceptron, the original data is subjected to denoising treatment through wavelet denoising, meanwhile, in order to further improve the fault diagnosis accuracy, the fault characteristics are better represented through data preprocessing, and the performance of a fault diagnosis model of the multi-layer perceptron is further improved. Therefore, the invention provides an algorithm combining wavelet change and a multi-layer perceptron for power transmission line fault identification, a fault feature matrix is constructed by wavelet change as the input of a multi-layer perceptron diagnosis model, an optimal diagnosis accuracy model is selected by adjusting the model structure and parameter adjustment, and the superiority of the multi-layer perceptron in power transmission line fault diagnosis is verified by comparing with the traditional machine learning.
The power transmission line fault diagnosis method is implemented according to the following steps based on the wavelet transformation-multilayer perceptron principle:
step 1, acquiring an original three-phase current signal of a power transmission system, and numbering fault types to determine fault type numbers corresponding to the original three-phase current signal; then, the original three-phase current signal is subjected to noise adding processing to obtain a fault sample containing noise, and the sample is divided into a training set and a testing set, wherein the specific process is as follows:
superposing a Gaussian white noise signal e (t) in an original three-phase current signal x (t) to obtain a current signal f (t) containing noise, wherein the function formula is as shown in the formula (1):
f(t)＝x(t)+e(t) (1)
in formula (1), t represents time, and the ratio of training set to test set in the examples is 4:1.
step 2, acquiring wavelet coefficients of different frequency bands through wavelet decomposition, determining a wavelet denoising threshold and a denoising mode to perform threshold denoising, and selecting a wavelet basis function with the best denoising effect to perform wavelet decomposition by verifying the denoising effect of different wavelet basis functions, wherein the specific process is as follows:
first, wavelet decomposition is adopted to obtain wavelet coefficients W on the kth frequency band on the jth layer j,k (t) wavelet coefficient W j,k The functional formula of (t) is shown as formula (2):
in the formula (2), a 0 j Is a telescoping factor ka 0 j b 0 As a translation factor, a 0 >1，b 0 >0; f (t) is a noisy current signalA number; psi is a wavelet basis function;
next, an adaptive threshold (Rigrsure) approach is used to determine the threshold lambda i The magnitude, i.e. by calculating the local mean mu i And local varianceThe threshold size is determined and the function is as follows:
wherein x (i) represents the value of the ith current signal and N represents the number of current value samples; according to local varianceDetermining a threshold lambda i When the magnitude is equal to or greater than k times sigma, the value of k is 2.5, the adaptive threshold lambda i The function of (2) is:
λ i ＝kσ i (5)
threshold lambda i After the determination, a hard threshold denoising mode and a soft threshold denoising mode which are commonly used in wavelet denoising are adopted, and the functions of the two modes are respectively shown as formulas (6) and (7):
wherein W is i Is the current wavelet coefficient vector; w (W) δ The wavelet coefficient is subjected to threshold value quantization;
finally, a denoising evaluation index is constructed, comprising a signal-to-noise ratio SNR (i.e., signal to Noise Ratio) and a root mean square error (i.e., RMSERoot Mean Square Error), wherein the two functions are respectively as follows:
wherein x (i) is the original signal; x is x δ (i) Is the denoised signal;is the power of the original signal;power for noise; n is the number of current value samples;
step 3, extracting wavelet coefficient energy values, constructing a feature matrix of fault energy, preprocessing the feature matrix in different modes, and selecting a data processing mode with highest diagnosis accuracy from the feature matrix, wherein the specific process is as follows:
by wavelet coefficients W j,k (t) obtaining energy E (j, k), wherein the energy E (j, k) has the following function formula:
the correlation between features can be reduced through standard normal distribution processing, and meanwhile, data of different orders of magnitude are converted into data of the same order of magnitude to facilitate comparison, and the standard normal distribution processing (namely, normalization processing) has the following functional formula:
wherein x' i Is the data after standardized processing; mu is the average value of the original data; sigma is the standard deviation of the original data;
step 4, establishing an MLP fault diagnosis model, adjusting the network full-connection layer structure and the number of neurons, adding Dropout network optimization, taking the one with the highest fault diagnosis accuracy as the MLP fault diagnosis model,
establishing a fault diagnosis model of the multi-layer perceptron, performing full-connection operation on all inputs through a plurality of neurons, extracting high-latitude fault characteristics, and outputting z through neurons of an activation function w,b The functional formula is shown as formula (13):
z w,b (x)＝ρ(w i x i +b) (13)
in formula (13), z w,b Representing the neuron output through the activation function, wherein ρ is the operation of the activation function;
output a of the nth layer of the multi-layer perceptron n It is required to satisfy the formula (14):
a n ＝ρ(z n )＝ρ(W n a n-1 +b n ) (14)
in the formula (14), a n Is z n Output, z, after operation by the activation function ρ n For the calculation result of the n-th layer of the multi-layer perceptron, W n Weight of the nth layer, b n Is the bias coefficient of the n-th layer;
calculating the output a of each layer by the formula (14) n The matrix expression is as shown in formula (15):
a n output for the nth layer; a, a n 1 ,a n 2 ,…,a n k The 1 st neuron calculated value, the 2 nd neuron calculated value, … and the nth layer output respectivelyk neuron calculations; w (w) k,n B is the weight value between each layer n 1 ,b n 2 ,…,b n k The 1 st neuron bias coefficient, the 2 nd neuron bias coefficient, … and the kth neuron bias coefficient of the nth layer respectively;
to avoid overfitting during the network training process, dropout techniques are therefore added and model training is evaluated using a mean square error (Mean Squared Error, MSE) loss function, the expression of which is as follows (16):
in the formula (16), w is a weight matrix; x is the network input; and y is network output.
And (3) experimental verification:
the invention relates to a power transmission line fault diagnosis method based on a wavelet-multilayer perceptron, wherein the constructed five-layer multilayer perceptron fault diagnosis model structure is shown in figure 1, the used Dropout network optimization is shown in figures 2a and 2b, the diagnosis flow is shown in figure 3, and the specific implementation steps are as follows:
step 1, acquiring an original three-phase current signal of a power transmission system, and numbering fault types to determine fault type numbers corresponding to the original three-phase current signal. Carrying out noise adding processing on an original three-phase current signal to obtain a fault sample containing noise, and carrying out sampling according to the following steps: 1, dividing the training set and the testing set;
and adding Gaussian white noise e (t) to the original three-phase current signal x (t) to obtain a current signal f (t) containing noise:
f(t)＝x(t)+e(t) (1)
step 2, first, wavelet decomposition is adopted to obtain wavelet coefficient W on the kth frequency band on the jth layer j,k (t) the functional formula is as shown in formula (2):
by calculating local mean mu i And local varianceDetermining a threshold lambda i The size, the function formula is:
the obtained threshold lambda i As formula (5):
λ i ＝kσ i (5)
k has a value of 2.5, after a threshold value is determined, a hard threshold value and a soft threshold value denoising mode are selected, and the functions are shown as formulas (6) and (7):
wherein W is i Is the current wavelet coefficient vector; w (W) δ The wavelet coefficient is obtained through threshold value quantization processing; lambda (lambda) i For a determined threshold size;
finally, constructing a denoising evaluation index signal-to-noise ratio and root mean square error to verify the denoising effect, wherein the functional formulas are shown as formula (8) and formula (9):
wherein, the liquid crystal display device comprises a liquid crystal display device,is the power of the original signal; />Power for noise; x (i) is the original signal; x is x δ (i) Is the denoised signal; n is the number of current value samples;
the sym6 wavelet basis function was selected for denoising, and the set evaluation index is shown in table 1.
TABLE 1 selection of sym6 wavelet basis function for denoising evaluation index
As is clear from analysis of the soft threshold and hard threshold denoising evaluation indexes, when the sym6 wavelet basis function is selected, the SNR value 45.3277 is the maximum value, and when the RMSE value is the minimum value, it is confirmed that the sym6 wavelet basis function denoising effect is the best, and therefore the sym6 wavelet basis function is selected for signal decomposition.
Step 3, selecting a wavelet basis function sym6 to perform signal 6-layer decomposition, extracting energy E (j, k) of each wavelet coefficient, wherein the function formula is as shown in formula (10):
wherein W is j,k (t) wavelet coefficients in a kth frequency band on a jth layer;
the original energy values of 7 frequency bands of faults are obtained according to the formula (10), and are shown in fig. 4a, 4b and 4c.
Preprocessing wavelet coefficient energy characteristics by adopting standard normal distribution, reducing correlation among the characteristics, simultaneously converting data of different orders of magnitude into the same order of magnitude, and performing standardized calculation as shown in the formula (11) and the formula (12):
x' i is data after standardized processing; mu is the average value of the original data; sigma is the standard deviation of the original data;
the normalized fault energy values are shown in fig. 5a, 5b and 5c.
Step 4, establishing an MLP fault diagnosis model, performing full-connection operation on all inputs through a plurality of nerves, extracting high latitude fault characteristics, and outputting z through neurons of an activation function w,b As in formula (13):
z w,b (x)＝ρ(w i x i +b) (13)
the input of the n-th layer satisfies the formula (14):
a n ＝ρ(z n )＝ρ(W n a n-1 +b n ) (14)
calculating the final layer output a by the method (14) n As formula (15):
the error function used in the network training process is a mean square error (Mean Squared Error, MSE) that evaluates the model training, the MSE is calculated as:
wherein the weight matrix is W, N is the number of samples, and x is the network input; y is network output;
in the fault simulation process of the present invention, 8 examples of tests were performed, and the fault simulation parameter settings of examples 1, 2, 3, 4, 5, 6, 7 and 8 are shown in table 2.
TABLE 2 Fault simulation parameter settings for 8 embodiments of the invention
Through the data of the 8 embodiments, the power transmission line fault diagnosis method based on the wavelet-multilayer perceptron has more reliable and stable effect on line fault monitoring. Fig. 6 is an equivalent circuit diagram of a power transmission line, wherein an electrical diagram is built in PSCAD according to a line structure, the power transmission line is a power transmission line with the total length of 5.0km and 10kV, a 10kV type YJLY-10kV crosslinked polyethylene power transmission line is selected for verification, parameters are shown in table 3, and a power transmission line fault PSCAD model is shown in fig. 7.
TABLE 3 Transmission line parameters
In order to increase the diversity of fault characteristics, the fault diversity is increased by modifying the time constant and the arc loss energy, the initial angle of occurrence of type faults is a random angle in one period, the fault distance is 0-5 km random position, sample data are sampled at 10kHz, each type of faults respectively generates 600 samples, and then the samples are processed according to 4: the scale of 1 is divided into training and validation sets. The result shows that the accuracy of the method for identifying the fault diagnosis of the power transmission line is obviously improved compared with the traditional machine learning Random Forest (RF), K Nearest Neighbor (KNN) and other methods.
Example 9 is a hardware test example of the present invention. The wavelet-multilayer perceptron of the MLP fault diagnosis model adopts 3 hidden layers, the number of neurons of the 3 hidden layers is 100-50-100 respectively, and a Dropout technology is added in the training process to avoid the overfitting phenomenon. In addition, different inputs are used as inputs of the multi-layer perceptron, dropout is set to be 0.4, learning rate is set to be 0.05, iteration is set to be 600, and the result of model training accuracy is shown in FIG. 8. As can be seen from fig. 8, the original current is taken as input, and the training accuracy fluctuates greatly due to the existence of redundant information in the initial training stage; the standardized energy is used as input, the training accuracy rate steadily rises, and after 250 iterations, the accuracy rate basically reaches the highest value, which is 98.72%; while the accuracy of the original current as input is still in large fluctuation after 600 iterations. The test set results are shown in fig. 9a and 9b, and 600 samples are divided into 480 training sets and 120 test sets, so that the fault identification accuracy reaches 98.39%.
Finally, to verify the advantages of the multi-layer perceptron adopted by the method, the feasibility of fault diagnosis of the method is further verified by comparing the traditional machine learning RF and K nearest neighbor methods and displaying the fault diagnosis result through the t-SNE visualization technology, wherein the diagnosis results are shown in fig. 10a, 10b and 10c, and XY represents the space position. Analyzing the fault diagnosis scatter diagram of each model, fig. 10a shows that the model test accuracy is 98.39%, and the scatter diagram shows that the feature vectors are divided into 8 obvious categories through calculation of multi-layer full-connection layers, the inter-fault distance of the same type is smaller, the distance between different features is larger, so that the MLP fault diagnosis model has less confusion between different faults in the fault recognition process, and the generalization of the MLP fault diagnosis model is stronger. Fig. 10b shows the RF classification result, the accuracy of the model test set is 93.75%, and since RF is composed of multiple DTs, the model identification accuracy is relatively improved by 4.51%, and as can be seen in the scatter diagram, the overall classification result of the model is better, wherein only the distance between the features of sample 4 and sample 5 is closer, and a small amount of confusion exists. Fig. 10c shows KNN classification results, the accuracy of the model test set is 89.24%, the effect is better in classification of sample 1, sample 6, sample 7 and sample 8, sample 2, sample 3 and sample 4 are divided into different sets, and there is partial overlap between sample 4 and sample 5. Through the comparison, a small amount of confusion exists between the sample 4 and the sample 5 in each traditional machine learning, and the comprehensive comparison shows that the MLP fault diagnosis model has stronger generalization in fault diagnosis.
Claims (5)
1. The power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron is characterized by comprising the following steps of: step 1, acquiring an original three-phase current signal, and numbering fault types; adding noise to an original three-phase current signal, obtaining a fault sample containing noise, and dividing the fault sample into a training set and a testing set; step 2, selecting a wavelet basis function with the best denoising effect to carry out wavelet decomposition; step 3, extracting wavelet coefficient energy values, constructing a feature matrix of fault energy, preprocessing the feature matrix in different modes, and selecting a data processing mode with highest diagnosis accuracy; and 4, adjusting the network full-connection layer structure and the number of neurons, adding Dropout network optimization, and taking the one with the highest fault diagnosis accuracy as an MLP fault diagnosis model.
2. The power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron as set forth in claim 1, wherein in step 1, the specific process is as follows:
collecting an original three-phase current signal of a power transmission system, and numbering fault types; then, carrying out noise adding processing on the original three-phase current signal to obtain a fault sample containing noise, and dividing the sample into a training set and a testing set;
superposing a Gaussian white noise signal e (t) in an original three-phase current signal x (t) to obtain a current signal f (t) containing noise, wherein the function formula is as shown in the formula (1):
f(t)＝x(t)+e(t)(1)
in formula (1), t represents time.
3. The power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron as set forth in claim 1, wherein in step 2, the specific process is as follows:
wavelet coefficients of different frequency bands are obtained through wavelet decomposition, a wavelet denoising threshold value and a denoising mode are determined to perform threshold denoising, a wavelet basis function with the best denoising effect is selected to perform wavelet decomposition,
first, wavelet decomposition is adopted to obtain wavelet coefficients W on the kth frequency band on the jth layer j,k (t) wavelet coefficient W j,k The functional formula of (t) is shown as formula (2):
in the formula (2), a 0 j Is a telescoping factor ka 0 j b 0 As a translation factor, a 0 >1，b 0 >0; f (t) is a noisy current signal; psi is a wavelet basis function;
next, an adaptive threshold mode is adopted to determine the threshold lambda i The magnitude, i.e. by calculating the local mean mu i And local varianceThe threshold size is determined and the function is as follows:
wherein x (i) represents the value of the ith current signal and N represents the number of current value samples; according to local varianceDetermining a threshold lambda i Size, then adaptive threshold lambda i The function of (2) is:
λ i ＝kσ i (5)
threshold lambda i After the determination, a hard threshold denoising mode and a soft threshold denoising mode which are commonly used in wavelet denoising are adopted, and the functions of the two modes are respectively shown as formulas (6) and (7):
wherein W is i Is the current wavelet coefficient vector; w (W) δ The wavelet coefficient is subjected to threshold value quantization;
finally, a denoising evaluation index is constructed, wherein the denoising evaluation index comprises a signal-to-noise ratio SNR and a root mean square error, and the two functions are respectively as follows:
wherein x (i) is the original signal; x is x δ (i) Is the denoised signal;is the power of the original signal; />Power for noise; n is the number of current value samples.
4. The power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron as set forth in claim 1, wherein in step 3, the specific process is as follows:
by wavelet coefficients W j,k (t) obtaining energy E (j, k), wherein the energy E (j, k) has the following function formula:
the correlation between features can be reduced through standard normal distribution processing, and meanwhile, data of different orders of magnitude are converted into data of the same order of magnitude to facilitate comparison, and the standard normal distribution processing has the following functional formula:
wherein x is i ' is normalized data; mu is the average value of the original data; sigma is the standard deviation of the original data.
5. The power transmission line fault diagnosis method based on the wavelet transformation-multilayer perceptron as set forth in claim 1, wherein in step 4, the specific process is as follows:
establishing a fault diagnosis model of the multi-layer perceptron, performing full-connection operation on all inputs through a plurality of neurons, extracting high-latitude fault characteristics, and outputting z through neurons of an activation function w,b The functional formula is shown as formula (13):
z w,b (x)＝ρ(w i x i +b) (13)
in formula (13), z w,b Representing the neuron output through the activation function, wherein ρ is the operation of the activation function;
output a of the nth layer of the multi-layer perceptron n It is required to satisfy the formula (14):
a n ＝ρ(z n )＝ρ(W n a n-1 +b n ) (14)
in the formula (14), a n Is z n Output, z, after operation by the activation function ρ n For the calculation result of the n-th layer of the multi-layer perceptron, W n Weight of the nth layer, b n Is the bias coefficient of the n-th layer;
calculating the output a of each layer by the formula (14) n The matrix expression is as shown in formula (15):
a n output for the nth layer; a, a n 1 ,a n 2 ,…,a n k The 1 st neuron calculated value, the 2 nd neuron calculated value, the … th neuron calculated value and the kth neuron calculated value output by the nth layer respectively; w (w) k,n B is the weight value between each layer n 1 ,b n 2 ,…,b n k The 1 st neuron bias coefficient, the 2 nd neuron bias coefficient, … and the kth neuron bias coefficient of the nth layer respectively;
in order to avoid overfitting in the network training process, a Dropout technology is added, and model training is evaluated by adopting a mean square error loss function, wherein the expression of the MSE loss function is as shown in a formula (16):
in the formula (16), w is a weight matrix; x is the network input; y is the network output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310538831.3A CN116559728A (en) | 2023-05-12 | 2023-05-12 | Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310538831.3A CN116559728A (en) | 2023-05-12 | 2023-05-12 | Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116559728A true CN116559728A (en) | 2023-08-08 |
Family
ID=87491250
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310538831.3A Pending CN116559728A (en) | 2023-05-12 | 2023-05-12 | Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116559728A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746323A (en) * | 2023-12-19 | 2024-03-22 | 国家电网有限公司 | Hydropower station auxiliary equipment fault detection method based on space-time state diagram |
-
2023
- 2023-05-12 CN CN202310538831.3A patent/CN116559728A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746323A (en) * | 2023-12-19 | 2024-03-22 | 国家电网有限公司 | Hydropower station auxiliary equipment fault detection method based on space-time state diagram |
| CN117746323B (en) * | 2023-12-19 | 2024-06-04 | 国网湖北省电力有限公司黄龙滩水力发电厂 | Hydropower station auxiliary equipment fault detection method based on space-time state diagram |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109213121B (en) | Method for diagnosing clamping cylinder fault of fan braking system | |
| CN111830408A (en) | Motor fault diagnosis system and method based on edge calculation and deep learning | |
| CN109034054B (en) | Harmonic multi-label classification method based on LSTM | |
| CN106548230A (en) | Diagnosis Method of Transformer Faults based on Modified particle swarm optimization neutral net | |
| CN111680788B (en) | Equipment fault diagnosis method based on deep learning | |
| CN114692665B (en) | Radiation source open set individual identification method based on metric learning | |
| CN111914705A (en) | Signal generation method and device for improving health state evaluation accuracy of reactor | |
| CN112327149A (en) | SSD-SRAE-based high-voltage circuit breaker fault diagnosis method | |
| CN111680875A (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
| CN116559728A (en) | Power transmission line fault diagnosis method based on wavelet transformation-multilayer perceptron | |
| CN110020637A (en) | A kind of analog circuit intermittent fault diagnostic method based on more granularities cascade forest | |
| CN112668105A (en) | Helicopter transmission shaft abnormity judgment method based on SAE and Mahalanobis distance | |
| CN115600088A (en) | Distribution transformer fault diagnosis method based on vibration signals | |
| Li et al. | Intelligent fault diagnosis of aeroengine sensors using improved pattern gradient spectrum entropy | |
| Jing et al. | A novel differentiable neural network architecture automatic search method for GIS partial discharge pattern recognition | |
| CN112486137A (en) | Method and system for constructing fault feature library of active power distribution network and fault diagnosis method | |
| CN111079647A (en) | Circuit breaker defect identification method | |
| Zhao et al. | Bearing fault diagnosis based on mel frequency cepstrum coefficient and deformable space-frequency attention network | |
| CN113611568A (en) | Vacuum circuit breaker based on genetic convolution depth network | |
| Li et al. | Transformer-based meta learning method for bearing fault identification under multiple small sample conditions | |
| CN117219124A (en) | Switch cabinet voiceprint fault detection method based on deep neural network | |
| CN112381667A (en) | Distribution network electrical topology identification method based on deep learning | |
| CN116340812A (en) | Transformer partial discharge fault mode identification method and system | |
| CN116029295A (en) | Electric power text entity extraction method, defect positioning method and fault diagnosis method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |