CN115048955A - Transformer vibration abnormity detection method based on time domain data segmentation - Google Patents
Transformer vibration abnormity detection method based on time domain data segmentation Download PDFInfo
- Publication number
- CN115048955A CN115048955A CN202210600954.0A CN202210600954A CN115048955A CN 115048955 A CN115048955 A CN 115048955A CN 202210600954 A CN202210600954 A CN 202210600954A CN 115048955 A CN115048955 A CN 115048955A
- Authority
- CN
- China
- Prior art keywords
- transformer
- vibration signal
- vibration
- neural network
- current
- 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
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H1/00—Measuring characteristics of vibrations in solids by using direct conduction to the detector
-
- 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
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Protection Of Transformers (AREA)
Abstract
The invention discloses a transformer vibration abnormity detection method based on time domain data segmentation. The technical scheme of the invention is as follows: dividing the operation working conditions of the transformer according to the load current and the operation voltage value of the transformer, and acquiring vibration signals of the surface of the transformer under different working conditions in normal operation by using an acceleration sensor; determining the shortest length of a vibration signal describing the state of the transformer based on the sampling frequency of the vibration signal and the vibration fundamental frequency of the transformer; and then determining the sliding step length and the optimal sample length of data segmentation through formula derivation and experiments, and extracting vibration signal characteristics. And finishing the classification of the current vibration signals of the transformer by using the trained one-dimensional convolution neural network, and judging that the transformer vibrates abnormally when the classification result is inconsistent with the current actual working condition. The method can obviously improve the feature extraction capability of the neural network, and has the advantages of simple segmentation and high detection precision.
Description
Technical Field
The invention relates to a transformer vibration abnormity detection method based on time domain data segmentation, and belongs to the field of electric power.
Background
The operation state of the transformer has an important influence on the safe and stable operation of the power grid. Theoretical research shows that the amplitude of the fundamental frequency of the vibration of the transformer iron core is in direct proportion to the magnitude of power voltage, and the amplitude of the fundamental frequency of the vibration of the winding is in direct proportion to the magnitude of load current, so that the working states of the winding and the iron core can be analyzed through the vibration signal of the transformer. The transformer vibration on-line monitoring has the advantages of being capable of being monitored in an electrified mode, low in cost, high in safety without direct electrical connection and the like, but the transformer surface vibration is influenced by multiple factors, the theory and method for analyzing the transformer operation state and diagnosing faults are far immature, and the actual requirements of transformer operation management and maintenance work are difficult to meet.
Disclosure of Invention
The invention provides a transformer vibration abnormity detection method based on time domain data segmentation. The technical scheme of the invention is as follows: dividing the operation condition of the transformer according to the load current and the operation voltage value of the transformer, and acquiring the surface vibration signals of the transformer under different working conditions in normal operation by using an acceleration sensor; determining the shortest length of a vibration signal describing the state of the transformer based on the sampling frequency of the vibration signal and the vibration fundamental frequency of the transformer; and then determining the sliding step length and the optimal sample length of data segmentation through formula derivation and experiments, and extracting vibration signal characteristics. And finishing the classification of the current vibration signals of the transformer by using the trained one-dimensional convolution neural network, and judging that the transformer vibrates abnormally when the classification result is inconsistent with the current actual working condition.
The invention has the advantages that:
(1) the segmentation method is simple and effective: the method determines the length of the transformer vibration signal sample based on the sampling frequency and the transformer vibration fundamental frequency analysis, and the segmentation method is simple and effective and does not need manual setting.
(2) The abnormality detection precision is high: the optimal length of the sample determined by the method comprises 2-3 vibration periods. Based on the sample setting, compared with other segmentation methods, the method disclosed by the patent can obviously improve the characteristic extraction capability of the convolutional neural network on the transformer vibration signal.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a transformer vibration anomaly detection method based on time domain data segmentation;
FIG. 2 is a schematic view of vibration data slicing;
FIG. 3 is a diagram of a modified one-dimensional convolutional neural network;
FIG. 4 activation function graph;
FIG. 5 is a schematic diagram of a one-dimensional convolution global average pooling operation;
FIG. 6 illustrates a Dropout operation;
FIG. 7 is another schematic diagram of a transformer vibration signal identification method based on time domain data segmentation;
FIG. 8 is a model training flow diagram;
fig. 9 compares different segmentation methods.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the present invention is further described with reference to the accompanying drawings:
as shown in fig. 1, the technical solution of the present invention is: dividing the operation working conditions of the transformer according to the load current and the operation voltage value of the transformer, and acquiring vibration signals of the surface of the transformer under different working conditions in normal operation by using an acceleration sensor; determining the shortest length of a vibration signal describing the state of the transformer based on the sampling frequency of the vibration signal and the vibration fundamental frequency of the transformer; and then determining the sliding step length and the optimal sample length of data segmentation through formula derivation and experiments, and extracting vibration signal characteristics. And finishing the classification of the current vibration signals of the transformer by using the trained one-dimensional convolution neural network, and judging that the transformer vibrates abnormally when the classification result is inconsistent with the current actual working condition.
1. Time domain data segmentation method
The vibration of the transformer mainly comes from the vibration of the iron core and the winding and has fixed frequency, the measured amplitude of the vibration signal on the surface of the running transformer also presents periodic transformation, and the vibration signal on the surface of the transformer can be approximately regarded as a stable signal. When the transformer has iron core or winding faults, only the amplitude of the fundamental frequency of the vibration signal changes, and the vibration period does not change. The graph of the slicing of the vibration data is shown in fig. 2.
Suppose the period of vibration T of the transformer surface 0 The number of sampling points in one period is the minimum length which can describe the state of the transformer and the length L of the sampling points is unchanged min The calculation formula of (2) is as follows:
L min ＝T 0 ·Fs (1)
in FIG. 1, W is the sliding step, N min Is the sample length after slicing. To ensure that the sample is cut sufficiently, the relation:
W≤L min ≤N min (2)
W≤L min the representation window is overlapped and moved, so that the waste of samples can be avoided, and the generation of enough samples is ensured. L is min ≤N min And the number of points of the experimental sample at least containing one vibration period of the transformer is shown, so that the cut sample is ensured to contain enough vibration information.
In data slicing, different data types are defined according to the difference of initial phases containing the vibration period length of the transformer in each sample. Phase difference of each sample
Comprises the following steps:
can calculate the passing N s After the samples with different phases, the 1 st sample and the Nth sample s If the initial phases of +1 samples are the same, the original signal will generate N after time domain slicing s Different types of data. Can calculate N s Comprises the following steps:
in the formula (3), N s Taking an integer.
When formula (3) is introduced into formula (4), it is possible to obtain:
the formula is combined to obtain:
to sum up, the transformer time domain vibration signal segmentation method provided by the patent comprises the following steps:
1) the sampling frequency of the known vibration acceleration sensor is FsHz, and the vibration period T of the transformer 0 The minimum length that can describe the transformer state is calculated from equation (1) as L min 。
2) Data acquisition by sliding window, sliding step W, L min And length of the test sample N min The three components need to satisfy the formula (2): w is less than or equal to L min ≤N min 。
3) And (4) calculating the sliding step length W of the experimental sample by the formula (6) to finish data segmentation.
2. Improved one-dimensional convolutional neural network
The improved one-dimensional convolutional neural network realized by the invention is obtained by improving on the basis of a classical LeNet-5 network, and the structure of the improved one-dimensional convolutional neural network is shown in figure 3.
The convolution layer uses different convolution check inputs to carry out convolution operation along the sequence advancing direction according to step length, and then a corresponding characteristic diagram can be obtained through an activation function. Due to the weight parameter sharing mechanism of the convolution kernels, each convolution kernel corresponds to one feature map, and the number of the convolution kernels is the depth of the output feature map. The convolution formula is shown in formula (7).
In the formula (7), the reaction mixture is,
is the output of the ith channel of the l-1 layer; cl-1 is the c channel of l-1 layer; y is l Is the output of the l-th layer;
a first layer convolution kernel weight matrix;
is a bias term; is a convolution operation.
The function of the activation function is to introduce a non-linear factor into the neural network to enhance the non-linear expressive power of the network. Commonly used saturation activation functions are Sigmoid, tanh and ReLU. The graphs of the three activation functions are shown in fig. 4. The mathematical expressions for the three activation functions are:
a l(i,j) ＝ReLU(x l(i,j) )＝max(0,x l(i,j) ) (10)
in the formula (8) -formula (10), x l(i,j) The jth characteristic value of the ith characteristic diagram in the ith convolution layer; a is l(i,j) Is x (i,j) An activation value of; a is a non-linear scale.
As can be seen from fig. 4, the derivatives of Sigmoid and tanh functions are close to 0 when approaching 0 and 1, and the neural network has a gradient vanishing phenomenon when propagating in the reverse direction. The ReLU function sets the characteristic value less than 0 to zero, which can accelerate the convergence speed of the model, and the activation function used in the patent is the ReLU function.
The role of the pooling layer is to compress the feature dimensions to reduce the amount of parameters for network operations, while reducing the likelihood of over-fitting. The patent adopts global average pooling, and the calculation process is shown in fig. 5.
In the neural network training process, under the condition that the training samples are too few, the model is easy to generate an overfitting phenomenon in the training process. To solve this problem, a Batch Normalization layer (BN) and Dropout are introduced into the patent model to improve its generalization capability. The BN layer normalizes the upper layer output, often placed between the convolutional layer and the activation function. The Dropout operation randomly discards a part of neurons according to a set probability in the training process, and the part of neurons only retains weight information. The robustness of the network is improved through multi-batch iteration. Fig. 6 is a schematic drawing of Dropout operation, in which the blue band "x" portion is the lost neuron.
The full connection layer is used for unfolding the extracted features of the previous layers into one-dimensional vectors, further extracting the features, and finally accessing a Softmax classifier to finish classification tasks.
The one-dimensional convolutional neural network model designed by the patent consists of an input layer, a convolutional layer I, a pooling layer I, a convolutional layer II, a pooling layer II, a convolutional layer III, a pooling layer III and two layers of full-connected layers, wherein a BN layer and a Drapout operation are introduced into the network, and the improvement of the classic CNN is as follows:
1) the input layer is one-dimensional original data subjected to time domain segmentation, and in order to adapt to the one-dimensional vibration time sequence array characteristic, a convolution kernel and a pooling kernel in the network both adopt one-dimensional structures;
2) convolutional layer I uses a larger size convolution kernel to increase the perceived field of view. The large convolution kernel is more beneficial to inhibiting high-frequency noise;
3) after the layers are wrapped, the model generalization capability is enhanced by the BN layer operating with Dropout.
Referring to fig. 7, the specific steps of applying the present invention are:
step 1: acquiring a vibration signal of the surface of the transformer by using an acceleration sensor;
the method comprises the steps of placing a vibration acceleration sensor at a preset vibration measuring point on the surface of the power transformer, setting the sampling rate of the vibration acceleration sensor to be Fs (10 kHz), setting the sampling time length of each time to be 0.5-1.0 second, setting the sampling interval to be 5-10 minutes, and recording the load current and voltage during each sampling.
Step 2: the method comprises the following steps of dividing operation condition intervals according to the load current and the operation voltage of the transformer:
1) and dividing the operation condition of the transformer according to the rated load current IN and the rated voltage UN of the transformer. The transformer is divided into four intervals of (0-60% IN ], (60% IN-95% IN ], (95% IN-105% IN ], (105% IN-130% IN ]) according to load current, and is divided into three intervals of (90% UN-98% UN ], (98% UN-102% UN ], (102% UN-120% UN) according to voltage, based on two factors of actual load current and voltage of the transformer, the operation condition of the transformer is divided into 12 types according to the interval division method, and the operation condition division is shown IN table 1:
TABLE 1 working condition division table
2) According to the transformer operation condition division method, a training data set based on condition division is constructed by applying vibration signal samples.
And step 3: performing time domain data segmentation on each type of transformer state, wherein the data segmentation pretreatment comprises the following steps:
1) the sampling frequency of the vibration acceleration sensor is FsHz, and the period of the vibration fundamental frequency (2 times of the power frequency) of the transformer is T 0 The shortest length that can describe the transformer state is calculated from equation (1):
L min ＝T 0 ·Fs (1)
the sampling frequency Fs is 10kHz, the voltage power frequency in China is 50Hz, and the current research conclusion shows that the surface vibration fundamental frequency of the transformer is twice of the power frequency, and the minimum length L capable of describing the state of the transformer can be calculated according to the formula (1) min ＝200。
2) Slicing of transformer surface vibration signal by sliding window, sliding step W, L min And sample length N min The three components need to satisfy the formula (2):
W≤L min ≤N min (2)
the sample length should satisfy equation (2): w is less than or equal to L min ≤N min I.e. the step size W should be equal to or less than 200 and the sample length should be equal to or greater than 200. Experimental analysis shows that the minimum sample length should at least contain 2-3L min The patent sets the sample length to 600.
3) And (4) calculating the sliding step length of the sample by the formula (3) to complete the segmentation of the vibration signal on the surface of the transformer, thereby obtaining the vibration signal sample. The number of different samples in equation (3) is calculated from equation (4).
And (4) calculating the sliding step length of the experimental sample through the formula (3) and the formula (4) to finish data segmentation. Experimental analysis shows that when N is present s When the length is less than or equal to 2, the feature extraction capability of the convolutional neural network is obviously enhanced, the comparison result is shown in fig. 9, the transformer vibration signal is segmented according to the method, and the optimal sample length and the optimal segmentation interval of the sample can be obtained.
It should be noted that the sequence between step 2 and step 3 may be interchanged, or the segmentation may be performed first, and then the working condition division may be performed; or the working condition division is carried out firstly and then the segmentation is carried out.
And 4, step 4: an improved one-dimensional convolutional neural network model is established, and a model training flow chart is shown in fig. 8. Training data and test data were as follows 4: 1, inputting the input of the one-dimensional convolution neural network model constructed by the method into a segmented transformer vibration signal, and outputting the input into a transformer operation condition state. The one-dimensional convolutional neural network model finishes the automatic extraction of state features through the alternation of convolutional layers and pooling layers. And finally, updating the network parameters through back propagation until the loss value meets the requirement and storing the model by taking the difference between the output probability of the Softmax function and the state class as a loss function.
The network hyper-parameter is set as: batch is 16, epoch is 30, and α is 0.001. The parameters of the network model satisfy table 1, the pooling layer is a global average pooling layer, the activation function is a ReLU function, and a Batch Normalization layer (BN) and Dropout are introduced into the model.
TABLE 1
And 5: and finally, applying the trained one-dimensional convolutional neural network anomaly detection model to carry out transformer vibration anomaly detection and judging the running state of the transformer. And preprocessing the data according to the model input requirement.
The method specifically comprises the following steps:
1) dividing the transformer state of on-line monitoring into normal and abnormal states, and sequentially dividing the transformer state into attention, alarm and fault according to the abnormal degree from low to high;
2) acquiring a vibration signal of the surface of a monitoring transformer by a vibration acceleration sensor to obtain a sample of the monitoring vibration signal, and determining the type of the actual working condition of the monitoring transformer according to the load current and the running voltage at the acquisition moment;
3) classifying the monitored vibration signal samples by using a one-dimensional convolutional neural network anomaly detection model, and judging the current state of the transformer according to the classification result and the difference of the actual working condition types, wherein the specific judgment rule is as follows:
the current working condition type can be calculated according to the collected voltage and current values, if the current working condition type is consistent with the detection result of the model, the current working condition type is normal, if the current working condition type is inconsistent with the detection result of the model, namely, if the detection result of the model is wrong, the current working condition type is considered to be an abnormal working condition, and the method specifically comprises the following steps:
(1) within 1 hour, when the interval of collecting and monitoring vibration signal samples is not less than 15 minutes every time and the condition that the classification result of the one-dimensional convolutional neural network anomaly detection model is inconsistent with the current actual working condition occurs for 2 times or more, judging that the current transformer running state is attention;
(2) within 3 hours, when the interval of collecting and monitoring vibration signal samples is not less than 30 minutes every time and the condition that the classification result of the one-dimensional convolution neural network anomaly detection model is inconsistent with the current actual working condition occurs for 4 times or more, judging that the current transformer running state is an alarm;
(3) when the interval of collecting and monitoring vibration signal samples is not less than 45 minutes every time within 6 hours, and the situation that the classification result of the one-dimensional convolutional neural network anomaly detection model is inconsistent with the current actual working condition occurs for 5 times or more, judging that the current transformer running state is a fault;
(4) if the 2 or all the conditions are met simultaneously, judging that the current running state is the state with the highest abnormal degree;
(5) and (3) when any one of the above items (1), (2) and (3) is not satisfied, judging that the current transformer running state is normal.
The invention discloses a transformer vibration abnormity detection method based on time domain data segmentation. The method comprises the following steps: and (3) performing segmentation pretreatment on time domain data, extracting the characteristics of samples containing 2-3 vibration periods after pretreatment, and detecting the vibration abnormality of the transformer through an improved one-dimensional convolution neural network. The transformer vibration abnormity detection method provided by the invention can improve the identification effect of the transformer vibration signal, effectively monitor the working state of the transformer and improve the working safety.
In the description, each part is described in a progressive manner, each part is emphasized to be different from other parts, and the same and similar parts among the parts are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (4)
1. A transformer vibration abnormity detection method based on time domain data segmentation is characterized by comprising the following steps: acquiring a transformer surface vibration signal by using an acceleration sensor, segmenting the transformer surface vibration signal, dividing a vibration signal sample according to working conditions to construct a training data set, and training and storing a one-dimensional convolution neural network anomaly detection model by using the training data set; finally, applying the trained one-dimensional convolutional neural network anomaly detection model to carry out transformer vibration anomaly detection;
wherein, use acceleration sensor to acquire transformer surface vibration signal, include: the vibration acceleration sensor is placed at the position where the vibration amplitude of the surface of the transformer is maximum, the sampling rate of the vibration acceleration sensor is set to be Fs (Fs is 10 k), the sampling time length of each time is 0.5-1.0 second, and the sampling interval is 5-10 minutes.
The method comprises the steps of segmenting a surface vibration signal of a transformer, obtaining a vibration signal sample, and setting the length N of the sample min Is 600, N s 2 or less, and the specific steps are as follows:
the method comprises the following steps: the sampling frequency of the vibration acceleration sensor is Fs, and the period of the vibration fundamental frequency (2 times of the power frequency) of the transformer is T 0 The shortest length that can describe the transformer state is calculated from equation (1):
L min ＝T 0 ·Fs (1)
step two: slicing of transformer surface vibration signal by sliding window, sliding step W, L min And sample length N min The three components need to satisfy the formula (2):
W≤L min ≤N min (2)
step three: calculating the sliding step length W of the sample by the formula (3), completing the segmentation of the surface vibration signal of the transformer, and obtaining a vibration signal sample, wherein N is in the formula (3) s The number of samples is calculated from equation (4).
2. The method according to claim 1, wherein the vibration signal samples are divided according to the working conditions to construct the training data set, and the specific steps are as follows:
the method comprises the following steps: dividing the operation condition of the transformer according to the load current and the operation voltage of the transformer;
according to rated load current I of transformer N And rated voltage U N And dividing the operation conditions of the transformer. According to the load current is divided into (0-60% I) N ]、(60％I N -95％I N ]、(95％I N -105％I N ]、(105％I N -130％I N ]Four intervals, divided by voltage (90% U) N -98％U N ]、(98％U N -102％U N ]、(102％U N -120％U N ) Three intervals. Based on two factors of the actual load current and the actual load voltage of the transformer, the operation working conditions of the transformer are divided into 12 types according to the interval division method.
Step two: according to the transformer operation condition division method, a training data set based on the condition division is constructed by applying the vibration signal samples.
3. The method of claim 2, wherein the one-dimensional convolutional neural network anomaly detection model is trained and stored using a training data set, as follows:
the input of the one-dimensional convolution neural network is a segmented vibration signal sample, and the output is the operation condition state of the transformer. The network hyper-parameter is set as: batch is 16, epoch is 30, and α is 0.001. The parameters of the network model satisfy table 1, the pooling layer is a global average pooling layer, the activation function is a ReLU function, and a Batch Normalization layer (BN) and Dropout are introduced into the model.
4. The method according to claim 3, wherein the transformer vibration anomaly detection is performed by applying a trained one-dimensional convolutional neural network anomaly detection model, and specifically comprises:
1) dividing the transformer state of on-line monitoring into normal and abnormal states, and sequentially dividing the transformer state into attention, alarm and fault according to the abnormal degree from low to high;
2) acquiring a vibration signal of the surface of a monitoring transformer by a vibration acceleration sensor, obtaining a sample of the monitoring vibration signal according to the methods of the claim 1, the claim 2 and the claim 3, and determining the type of the actual working condition of the monitoring vibration signal according to the load current and the operating voltage at the acquisition moment;
3) classifying the monitored vibration signal samples by using a one-dimensional convolutional neural network anomaly detection model, and judging the current state of the transformer according to the classification result and the difference of the actual working condition types, wherein the specific judgment rule is as follows:
(1) within 1 hour, when the interval of collecting and monitoring vibration signal samples is not less than 15 minutes every time and the condition that the classification result of the one-dimensional convolutional neural network anomaly detection model is inconsistent with the current actual working condition occurs for 2 times or more, judging that the current transformer running state is attention;
(2) within 3 hours, when the interval of collecting and monitoring vibration signal samples is not less than 30 minutes every time and the condition that the classification result of the one-dimensional convolution neural network anomaly detection model is inconsistent with the current actual working condition occurs for 4 times or more, judging that the current transformer running state is an alarm;
(3) when the interval of collecting and monitoring vibration signal samples is not less than 45 minutes every time within 6 hours, and the situation that the classification result of the one-dimensional convolutional neural network anomaly detection model is inconsistent with the current actual working condition occurs for 5 times or more, judging that the current transformer running state is a fault;
(4) if the 2 or all the conditions are met simultaneously, judging that the current running state is the state with the highest abnormal degree;
(5) and (3) when any one of the above items (1), (2) and (3) is not satisfied, judging that the current transformer running state is normal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210600954.0A CN115048955A (en) | 2022-05-30 | 2022-05-30 | Transformer vibration abnormity detection method based on time domain data segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210600954.0A CN115048955A (en) | 2022-05-30 | 2022-05-30 | Transformer vibration abnormity detection method based on time domain data segmentation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115048955A true CN115048955A (en) | 2022-09-13 |
Family
ID=83159051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210600954.0A Pending CN115048955A (en) | 2022-05-30 | 2022-05-30 | Transformer vibration abnormity detection method based on time domain data segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115048955A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150269A (en) * | 2023-09-04 | 2023-12-01 | 淮阴师范学院 | Electrical equipment operation abnormality diagnosis system based on data analysis |
| CN117150269B (en) * | 2023-09-04 | 2024-06-07 | 淮阴师范学院 | Electrical equipment operation abnormality diagnosis system based on data analysis |
-
2022
- 2022-05-30 CN CN202210600954.0A patent/CN115048955A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150269A (en) * | 2023-09-04 | 2023-12-01 | 淮阴师范学院 | Electrical equipment operation abnormality diagnosis system based on data analysis |
| CN117150269B (en) * | 2023-09-04 | 2024-06-07 | 淮阴师范学院 | Electrical equipment operation abnormality diagnosis system based on data analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1234092C (en) | Predictive modelling method application to computer-aided medical diagnosis | |
| CN108022014B (en) | Power system load prediction method and system | |
| CN110044623B (en) | Intelligent rolling bearing fault identification method based on empirical mode decomposition residual signal characteristics | |
| US20230316051A1 (en) | Pre-alarming method for rotary stall of compressors based on temporal dilated convolutional neural network | |
| CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
| CN113848417A (en) | Rail transit power supply equipment fault prediction method and device | |
| CN110263944A (en) | A kind of multivariable failure prediction method and device | |
| CN116628616B (en) | Data processing method and system for high-power charging energy | |
| Deng et al. | A deep learning method based on long short term memory and sliding time window for type recognition and time location of power quality disturbance | |
| CN111291918B (en) | Rotating machine degradation trend prediction method based on stationary subspace exogenous vector autoregression | |
| CN106533361A (en) | Wavelet analysis-based photovoltaic module fault diagnosis method | |
| CN115166514A (en) | Motor fault identification method and system based on self-adaptive spectrum segmentation and denoising | |
| CN114563671A (en) | High-voltage cable partial discharge diagnosis method based on CNN-LSTM-Attention neural network | |
| CN112798888B (en) | Non-invasive fault diagnosis method for vehicle-mounted electrical system of unmanned train | |
| CN108415880B (en) | Line loss characteristic analysis method based on sample entropy and wavelet transformation | |
| CN115048955A (en) | Transformer vibration abnormity detection method based on time domain data segmentation | |
| Liu et al. | Fault diagnosis method of rolling bearing based on the multiple features of LMD and random forest | |
| CN113159088A (en) | Fault monitoring and diagnosis method based on multi-feature fusion and width learning | |
| CN115169418A (en) | Switch cabinet partial discharge identification method and system based on improved wavelet threshold method | |
| CN116773952A (en) | Transformer voiceprint signal fault diagnosis method and system | |
| CN116566061A (en) | Grid-connected inverter system stability on-line monitoring method and system | |
| CN110866840A (en) | Database modeling method for power load characteristic quantity training based on knowledge graph | |
| CN116628620A (en) | Non-invasive load identification calculation method | |
| CN114609435A (en) | Voltage sag detection and classification identification method | |
| CN114818813A (en) | Cutter wear value prediction algorithm based on self-attention convolution neural network |
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 |