Power transformer state monitoring method and device

Abstract

The invention relates to the technical field of power equipment state monitoring, and discloses a power transformer state monitoring method, device and equipment and a computer program product. The method comprises the steps that sound signals in the operation process of a transformer are collected, and after windowing processing, self-adaptive cochlea filtering, wavelet transformation and normalization processing are carried out, the operation state of the transformer is analyzed based on an information entropy weighted grey relational degree algorithm. Compared with the limitation of only depending on a traditional signal processing method in the prior art, and particularly under a strong noise environment or a complex signal condition, the technical problem that fault judgment on the state of the transformer cannot be realized is solved, due to a processing mode of combining adaptive filtering and multi-resolution wavelet transform, the fault judgment on the state of the transformer can be realized; according to the invention, high-precision analysis of the sound signal of the transformer can be realized, so that the problem of misjudgment caused by noise interference is avoided, and the accuracy and real-time performance of fault detection are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment status monitoring, and in particular relates to a method and device for monitoring the status of a power transformer. Background Art

[0002] At present, as the core equipment in the power system, the operating status of the power transformer is directly related to the safety and stability of the power grid. However, traditional monitoring methods mainly rely on the detection of electrical parameters such as current, voltage, temperature and other signals, or process sound signals based on spectrum analysis methods such as Fourier transform and short-time Fourier transform. These methods can play a certain role when the working environment of the transformer is relatively stable and the noise interference is less, but in the face of more complex operating conditions such as strong noise environment, non-stationary signal or multiple interference source scenarios, traditional methods cannot effectively identify the potential early faults of the transformer, which often leads to false alarms and missed alarms. Especially when the transformer is in a high noise interference environment, the monitoring system cannot effectively distinguish between the external environmental noise and the weak fault signal inside the transformer, resulting in the inability to accurately extract key information. Therefore, there is an urgent need for a method that can still accurately extract key sound signals at different frequencies and time scales during the operation of the transformer in a complex noise environment and under multiple interference conditions, reduce the misjudgment rate, and ensure timely warning of early faults, thereby improving the operation safety and reliability of the power system. Summary of the invention

[0003] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to propose a method for monitoring the state of a power transformer, aiming to solve the limitations of the prior art that only relies on traditional signal processing methods, especially in a strong noise environment or complex signal conditions, and cannot realize the technical problem of fault judgment of the transformer state.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solution: The present invention provides a method for monitoring the state of a power transformer.

[0005] The method comprises:

[0006] Step S10: setting the sampling frequency and sampling duration of the transformer sound signal, the sampling frequency is 51200 Hz, the sampling duration is 2 seconds, and collecting the first sound signal of the transformer through the sound collection sensor, the first sound signal includes the low-frequency and high-frequency sound signals during the operation of the transformer;

[0007] Step S20: performing windowing processing on the first sound signal, wherein the windowing processing uses a Hanning window to perform segmented processing on the sound signal, wherein the length of each signal segment is 12288 sampling points, and the overlapping length of adjacent signal segments is 4096 sampling points, and each signal segment is subjected to one-dimensional compression processing to generate a one-dimensional compressed signal sequence;

[0008] Step S30: designing an adaptive cochlear filter bank, wherein the filter bank includes 18 filters, the center frequency of each filter matches the frequency spectrum distribution of the transformer sound signal, and the bandwidth range of the filter is 50 Hz to 1200 Hz. The one-dimensional compressed signal sequence is filtered by the filter bank to generate a cochlear characteristic coefficient matrix for each frequency band;

[0009] Step S40: applying a multi-resolution analysis method of wavelet transform to the cochlear characteristic coefficient matrix, decomposing the one-dimensional compressed signal sequence into frequency sub-bands of different scales, extracting multi-resolution feature vectors by analyzing the features at different scales, and fusing them to generate an auditory feature parameter matrix;

[0010] Step S50: Based on the mean and variance of the feature matrix, the auditory feature parameter matrix is ​​normalized to eliminate signal differences caused by different distances and intensities, thereby obtaining a normalized auditory feature matrix;

[0011] Step S60: Obtain the standard auditory feature parameter matrix of the transformer under normal operating conditions, and calculate the information entropy weighted grey correlation between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, including weighting the feature parameters according to the entropy weights in the frequency domain and the time domain, and the formula is:

[0012]

[0013] Among them, Δ i represents the absolute difference between the standard auditory feature parameter matrix and the auditory feature parameter matrix to be tested in i features, Δ min Represents the minimum value of all feature difference values, Δ max represents the maximum value among all feature difference values, ρ is the information entropy weighted grey correlation degree between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, 0<ρ<1, E f It represents the entropy weight in the frequency domain, which represents the importance of the transformer sound signal at different frequency components. t It represents the entropy weight in the time domain, reflecting the importance of the features of the transformer sound signal in different time periods, and n is the total number of features involved in the correlation calculation;

[0014] Step S70: If the value of the information entropy weighted grey correlation degree is greater than or equal to 0.8, the transformer operation state is determined to be normal; if it is less than 0.8, the transformer operation state is determined to be abnormal.

[0015] Preferably, in step S10, the sound collection sensor is an electromagnetic sensor or a piezoelectric sensor, which can capture sound signals of different frequency ranges generated during the operation of the transformer.

[0016] Preferably, in step S10, the sampling frequency and the sampling duration are adjusted according to the operating environment and type of different transformers.

[0017] Preferably, in step S20, the window function used in the windowing process further includes a Gaussian window.

[0018] Preferably, in step S30, the design of the adaptive cochlear filter bank can dynamically adjust the bandwidth and center frequency of each filter according to the frequency spectrum changes of the sound signal collected in real time.

[0019] Preferably, in step S70, when the information entropy weighted grey correlation value is less than 0.8, it is determined that the transformer operation state is abnormal and an audible and visual alarm device is triggered.

[0020] Preferably, in step S50, the normalization process includes using a Z-score normalization method.

[0021] The present invention also provides a device, comprising:

[0022] A signal sampling module is used to set the sampling frequency and sampling duration of the transformer sound signal, the sampling frequency is 51200 Hz, the sampling duration is 2 seconds, and the first sound signal of the transformer is collected by the sound collection sensor, the first sound signal includes the low-frequency and high-frequency sound signals during the operation of the transformer;

[0023] A signal processing module, used to perform windowing processing on the first sound signal, wherein the windowing processing uses a Hanning window to perform segmented processing on the sound signal, the length of each signal segment is 12288 sampling points, the overlapping length of adjacent signal segments is 4096 sampling points, and each signal segment is subjected to one-dimensional compression processing to generate a one-dimensional compressed signal sequence;

[0024] A filter design module is used to design an adaptive cochlear filter bank, wherein the filter bank includes 18 filters, the center frequency of each filter matches the frequency spectrum distribution of the transformer sound signal, and the bandwidth range of the filter is 50 Hz to 1200 Hz. The one-dimensional compressed signal sequence is filtered by the filter bank to generate a cochlear characteristic coefficient matrix for each frequency band;

[0025] A parameter matrix generation module is used to apply a multi-resolution analysis method of wavelet transform to the cochlear characteristic coefficient matrix, decompose the one-dimensional compressed signal sequence into frequency sub-bands of different scales, extract multi-resolution feature vectors by analyzing the features at different scales, and fuse them to generate an auditory feature parameter matrix;

[0026] A normalization processing module, used for normalizing the auditory feature parameter matrix based on the mean and variance of the feature matrix to eliminate signal differences caused by different distances and intensities, and obtaining a normalized auditory feature matrix;

[0027] The information entropy weighted grey correlation calculation module is used to obtain the standard auditory feature parameter matrix of the transformer under normal operating conditions, and calculate the information entropy weighted grey correlation between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, including weighting the feature parameters according to the entropy weights in the frequency domain and time domain. The formula is:

[0028]

[0029] Among them, Δ i represents the absolute difference between the standard auditory feature parameter matrix and the auditory feature parameter matrix to be tested in i features, Δ min Represents the minimum value of all feature difference values, Δ max represents the maximum value among all feature difference values, ρ is the information entropy weighted grey correlation degree between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, 0<ρ<1, E f It represents the entropy weight in the frequency domain, which represents the importance of the transformer sound signal at different frequency components. t It represents the entropy weight in the time domain, reflecting the importance of the features of the transformer sound signal in different time periods, and n is the total number of features involved in the correlation calculation;

[0030] The fault judgment module is used to judge that if the value of the information entropy weighted grey correlation degree is greater than or equal to 0.8, the transformer operation state is judged to be normal; if it is less than 0.8, the transformer operation state is judged to be abnormal.

[0031] The present invention also provides a power transformer state monitoring device, comprising:

[0032] A memory, a processor, and a power transformer state monitoring program stored in the memory and executable on the processor, wherein the power transformer state monitoring program implements the power transformer state monitoring method when executed by the processor.

[0033] The present invention also provides a computer program product, comprising a power transformer state monitoring program, wherein the power transformer state monitoring program implements the power transformer state monitoring method when executed by a processor.

[0034] The beneficial effect of the present invention is that compared with the prior art which only relies on the limitations of traditional signal processing methods, especially in strong noise environments or complex signal conditions, and cannot realize the technical problem of fault judgment of transformer status, the present application uses a processing method that combines adaptive filtering with multi-resolution wavelet transform to achieve high-precision analysis of transformer sound signals, thereby avoiding the problem of misjudgment caused by noise interference and improving the accuracy and real-time performance of fault detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 The figure is a flow chart of a first embodiment of a method for monitoring the state of a power transformer according to the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Embodiment 1: Figure 1 FIG. 1 is a flow chart of a first embodiment of a method for monitoring the state of a power transformer according to the present invention, and provides a first embodiment of a method for monitoring the state of a power transformer according to the present invention.

[0039] In a first embodiment, the method comprises:

[0040] Step S10: setting the sampling frequency and sampling duration of the transformer sound signal, the sampling frequency is 51200 Hz, the sampling duration is 2 seconds, and collecting the first sound signal of the transformer through the sound collection sensor, the first sound signal includes the low-frequency and high-frequency sound signals during the operation of the transformer;

[0041] It should be noted that during the operation of the transformer, the sound signal contains sound wave components of different frequency ranges, among which the low-frequency signal usually reflects the mechanical vibration characteristics of the transformer core and winding, while the high-frequency signal may be related to fault characteristics such as electrical discharge and partial discharge inside the transformer. The collected sound signal can fully cover all kinds of information on the operating status of the transformer, which is convenient for subsequent signal processing and feature extraction.

[0042] It should be understood that in order to better improve the acquisition accuracy of the sound signal, in this embodiment, the selected sampling frequency is 51200Hz. This sampling frequency can fully capture the low-frequency vibration and high-frequency fault signals during the operation of the transformer, ensuring that important fault information is not lost. At the same time, the sampling time is set to 2 seconds. This time can avoid excessive data volume while ensuring sufficient signal capture and reduce the complexity of subsequent processing. After the signal is acquired, it will be windowed to eliminate the edge effect caused by signal segmentation and improve the stability of feature extraction. In addition, the use of different types of sound acquisition sensors (such as electromagnetic sensors or piezoelectric sensors) can further optimize the quality of signal acquisition to adapt to the specific application scenarios and operating environments of different transformers.

[0043] Step S20: performing windowing processing on the first sound signal, wherein the windowing processing uses a Hanning window to perform segmented processing on the sound signal, wherein the length of each signal segment is 12288 sampling points, and the overlapping length of adjacent signal segments is 4096 sampling points, and each signal segment is subjected to one-dimensional compression processing to generate a one-dimensional compressed signal sequence;

[0044] It should be noted that the Hanning window is a commonly used windowing function, which is used to reduce the spectrum leakage problem when the signal is segmented. In the signal processing process, directly segmenting the unwindowed signal will lead to discontinuous jumps, thereby introducing spectrum leakage effects. By using the Hanning window to perform windowing on the signal, the signal boundary can be effectively smoothed, the noise interference caused by the edge effect can be reduced, and the accuracy of frequency domain analysis can be improved.

[0045] It should be understood that the selection of 12288 sampling points as the length of each signal segment is mainly based on the analysis of the signal frequency distribution. This number of sampling points can fully cover the main characteristic frequencies during the operation of the transformer, so that subtle changes in the sound signal can be effectively captured during feature extraction. At the same time, the overlapping length setting of 4096 sampling points can ensure that adjacent segmented signals have good continuity, avoid missing some key information due to a single sampling, and ensure that subsequent signal processing is more stable and accurate. In addition, in order to further reduce the amount of data and improve computational efficiency, a one-dimensional compression processing method is adopted to compress the segmented signal into a one-dimensional signal sequence. Through one-dimensional compression processing, the main characteristic information of the signal can be retained, while reducing data redundancy, making subsequent filtering and feature extraction more efficient.

[0046] It should be understood that the length and overlap of the segmented signals can be adjusted according to different application scenarios. For example, in a high noise environment, the length of the signal segment can be appropriately shortened and the overlap length can be increased to ensure that the key features in the signal are not masked by the noise, thereby improving the robustness of the system.

[0047] Step S30: designing an adaptive cochlear filter bank, wherein the filter bank includes 18 filters, the center frequency of each filter matches the frequency spectrum distribution of the transformer sound signal, and the bandwidth range of the filter is 50 Hz to 1200 Hz. The one-dimensional compressed signal sequence is filtered by the filter bank to generate a cochlear characteristic coefficient matrix for each frequency band;

[0048] It should be noted that the cochlear filter bank is a filter bank designed to simulate the auditory characteristics of the human cochlea, which can decompose broadband sound signals into several different frequency sub-bands to achieve feature extraction of signals in different frequency bands. Each filter corresponds to a specific frequency band. By adaptively adjusting the center frequency and bandwidth of the filter, it can effectively match the sound signals of different frequency components during the operation of the transformer, thereby capturing low-frequency mechanical vibration signals and high-frequency electrical fault signals.

[0049] It should be understood that the selection of 18 filters is based on the analysis of the distribution characteristics of the transformer sound signal spectrum. The filter bank can accurately cover the frequency bands from low frequency to high frequency (50Hz to 1200Hz), where the low frequency part mainly contains the mechanical vibration signal of the transformer, while the high frequency part may contain fault signals such as partial discharge. Through the decomposed signals of different frequency bands, the key features related to the transformer status can be effectively extracted.

[0050] It should be understood that the adaptive cochlear filter bank can automatically adjust the center frequency and bandwidth of each filter according to the collected sound signal, so as to adapt to the spectrum changes of the transformer sound signal under different operating conditions. This adaptive characteristic can further enhance the monitoring capability of the system in complex environments, ensuring that key signal features can still be accurately captured under conditions such as transformer load fluctuations and environmental noise interference.

[0051] Step S40: applying a multi-resolution analysis method of wavelet transform to the cochlear characteristic coefficient matrix, decomposing the one-dimensional compressed signal sequence into frequency sub-bands of different scales, extracting multi-resolution feature vectors by analyzing the features at different scales, and fusing them to generate an auditory feature parameter matrix;

[0052] It should be noted that wavelet transform is a signal processing tool that can provide information in both time domain and frequency domain, and is particularly suitable for processing non-stationary signals. The sound signal generated by the transformer during operation usually exhibits non-stationary characteristics, that is, the frequency components of the signal change continuously over time. Through multi-resolution analysis of wavelet transform, the signal can be decomposed into multiple sub-bands of different scales (frequency resolution and time resolution), thereby capturing detailed information of different frequency bands in the signal.

[0053] It should be understood that although the traditional Fourier transform can provide frequency information, it is essentially a global analysis and cannot provide sufficient resolution in both the time and frequency domains, especially when dealing with non-stationary signals in complex environments such as transformers, and cannot effectively capture local fault features. Compared with the Fourier transform, the wavelet transform in the present invention provides a multi-scale decomposition capability of local features in the time and frequency domains, can capture local changes in short-time high-frequency signals, and is particularly suitable for detecting early fault signals of transformers.

[0054] For example, assume that the data of the transformer sound signal collected is as follows:

[0055] Sampling frequency: 51,200Hz

[0056] Sampling time: 2 seconds

[0057] Number of sampling points: 102,400 (51200Hz×2 seconds)

[0058] After preliminary processing by the cochlear filter bank, a cochlear characteristic coefficient matrix X(t) is generated with a matrix size of 102,400×18, which represents the processing results of 102,400 sampling points through 18 filters.

[0059] (1) Daubechies 4 (db4) wavelet is selected as the decomposition wavelet basis function. This wavelet basis can provide higher time and frequency resolution and is suitable for processing non-stationary signals during transformer operation. In order to analyze the signal characteristics from low frequency to high frequency, a 6-layer wavelet decomposition is performed. This number of layers can effectively decompose different frequency bands during transformer operation.

[0060] (2) Perform a 6-layer wavelet decomposition on the cochlear characteristic coefficient matrix X(t) to obtain detail coefficients D at different scales j (t) and the approximation coefficient A j (t) The formula is as follows: Among them, A6(t) is the approximation coefficient of the 6th layer, which mainly contains low-frequency features, and D j (t) is the detail coefficient of the jth layer, which contains the high-frequency information of each layer. Through this multi-level decomposition, the characteristics of different frequency bands in the signal can be captured.

[0061] (3) Frequency distribution of signals in each layer:

[0062] Layer 1: 10,240Hz-25,600Hz (high frequency signal, usually used to detect partial discharge or short-duration burst fault signals)

[0063] Layer 2: 5,120Hz-10,240Hz (high frequency signal)

[0064] Layer 3: 2,560Hz-5,120Hz (medium and high frequency signals)

[0065] Layer 4: 1,280Hz-2,560Hz (intermediate frequency signal, may contain mechanical vibration information)

[0066] Layer 5: 640Hz-1,280Hz (low frequency signal)

[0067] Layer 6: 0Hz-640Hz (very low frequency fundamental signal, usually reflects the stable operation state of the transformer)

[0068] (4) By calculating the detail coefficient D at each scale j (t) and the energy value of the approximation coefficient A6(t) to extract features. The formula is:

[0069]

[0070]

[0071] Among them, represents the energy of the jth layer, representing the signal strength of this frequency band. For example, E1 = 150, the energy of the high-frequency signal is relatively high, which may indicate the presence of a partial discharge fault, E4 = 80, the energy of the intermediate frequency signal is low, indicating that the mechanical vibration is relatively stable, and E6 = 300, the low-frequency fundamental wave energy is high, reflecting the normal operation of the transformer.

[0072] (5) For each layer of signal energy, the auditory feature parameter matrix Y(t) is generated through the adaptive weighted fusion method, and the formula is as follows:

[0073]

[0074] Among them, the weight coefficient α j The energy E of each layer j Adaptive Adjustment:

[0075]

[0076] Assume that the weights are:

[0077] α1=0.3(high-frequency partial discharge signal is more significant)

[0078] α4=0.15(intermediate frequency vibration signal is weak)

[0079] α6=0.55 (low frequency signal is strong, reflecting normal operation)

[0080] The feature vectors at each level are weightedly fused using these weights, and finally the fused auditory feature parameter matrix Y(t) is obtained.

[0081] Step S50: Based on the mean and variance of the feature matrix, the auditory feature parameter matrix is ​​normalized to eliminate signal differences caused by different distances and intensities, thereby obtaining a normalized auditory feature matrix;

[0082] It should be noted that in the process of collecting transformer sound signals, different sensor installation positions and different sampling conditions (such as the distance between the sensor and the transformer, transformer workload, etc.) will lead to differences in signal strength. In order to solve these non-essential signal strength differences caused by the physical environment, the use of normalization processing can ensure that the feature matrix has a consistent signal representation under different operating environments. By normalizing the mean and variance of the feature matrix, the differences in signal amplitude can be effectively eliminated, ensuring the reliability of subsequent analysis and feature extraction.

[0083] It should be understood that in order to better eliminate these intensity differences and ensure the comparability of eigenvalues, the specific process of normalization can be calculated based on the mean and variance of the feature matrix. This method makes each feature have zero mean and unit variance, thereby unifying the scale of different features.

[0084] Step S60: Obtain the standard auditory feature parameter matrix of the transformer under normal operating conditions, and calculate the information entropy weighted grey correlation between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, including weighting the feature parameters according to the entropy weights in the frequency domain and the time domain, and the formula is:

[0085]

[0086] Among them, Δ i represents the absolute difference between the standard auditory feature parameter matrix and the auditory feature parameter matrix to be tested in i features, Δ min Represents the minimum value of all feature difference values, Δ max represents the maximum value among all feature difference values, ρ is the information entropy weighted grey correlation degree between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, 0<ρ<1, E f It represents the entropy weight in the frequency domain, which represents the importance of the transformer sound signal at different frequency components. t It represents the entropy weight in the time domain, reflecting the importance of the features of the transformer sound signal in different time periods, and n is the total number of features involved in the correlation calculation;

[0087] It should be understood that in order to improve the accuracy of fault detection, the frequency domain and time domain weighting mechanisms are introduced into the information entropy weighted grey correlation calculation, which are dynamically adjusted by the spectral distribution and time distribution of the signal respectively. This weighting method can highlight the frequency bands and time periods with rich fault feature information, making the monitoring of transformers under different operating conditions more accurate.

[0088] For example, in a certain monitoring, the signal difference between the standard feature matrix and the feature matrix to be tested at a frequency of 100Hz is Δ1=0.2, and the difference at a frequency of 200Hz is Δ2=0.05. Then for each feature, we can calculate the grey correlation coefficient.

[0089] Step S70: If the value of the information entropy weighted grey correlation degree is greater than or equal to 0.8, the transformer operation state is determined to be normal; if it is less than 0.8, the transformer operation state is determined to be abnormal.

[0090] It should be noted that the calculation of the gray correlation value γ is based on the similarity between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix. When γ≥0.8, it means that the operating state of the transformer to be tested has a high similarity with the normal operating state and belongs to the normal operating state; when γ<0.8, it means that there is a large difference between the state to be tested and the standard state, which may be due to some potential fault in the equipment.

[0091] It should be understood that the threshold of 0.8 was determined through multiple experiments and data analysis. This value is set as a compromise value, which can not only ensure that there will be no false alarms for smaller signal differences, but also detect and alarm major fault signals in a timely and accurate manner. In order to further improve the accuracy and robustness of diagnosis, the threshold can be appropriately adjusted based on the actual application scenario. For example: in an environment with large noise interference, the threshold may be appropriately increased to 0.85 to reduce false alarms; in sensitive application scenarios (such as when there is accumulation of small fault signals when the transformer works for a long time), the threshold can be reduced to 0.75 for early warning.

[0092] In the second embodiment provided by the present invention, in step S10, the sampling frequency and the sampling duration are adjusted according to the operating environment and type of different transformers;

[0093] It should be noted that the operating environment and type of the transformer have an important impact on the sampling frequency and sampling duration of the sound signal. The sound signals generated by different transformers have different frequency characteristics when working. For example, a high-power transformer may generate a lower-frequency mechanical vibration signal, while a smaller power or high-load transformer may generate a high-frequency fault signal (such as a partial discharge signal). Therefore, reasonably adjusting the sampling frequency and duration can more effectively capture the key signals related to transformer faults.

[0094] It should be understood that in order to better adapt to different types of transformers and their working environments, the sampling frequency and sampling duration can be flexibly adjusted according to the needs of actual applications. Specifically, the selection of the sampling frequency needs to be combined with the power size of the transformer and the type of key signals to be monitored. For large-power transformers or scenarios where low-frequency mechanical vibrations are mainly monitored, the sampling frequency is usually set between 10,000Hz and 20,000Hz, so that the mechanical vibration characteristic signals of the transformer can be fully captured. For transformers that may generate high-frequency fault signals, especially partial discharge or other types of electrical faults, the sampling frequency should be appropriately increased and set to 40,000Hz to 51,200Hz to ensure that high-frequency detail signals can be accurately captured. For example, when detecting a large power transformer, if the main monitoring is low-frequency vibration, the sampling frequency can be set to 15,000Hz; but when the system detects an electrical discharge fault, the sampling frequency needs to be increased to 51,200Hz to capture high-frequency fault signals.

[0095] The present invention provides a power transformer state monitoring device, which adopts a power transformer state monitoring method in the above embodiment, and can solve a technical problem of power transformer state monitoring. Compared with the prior art, the beneficial effects of the power transformer state monitoring device provided by the present invention are the same as the beneficial effects of the power transformer state monitoring method provided by the above embodiment, and other technical features in the power transformer state monitoring device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0096] The present invention provides a power transformer state monitoring device, the power transformer state monitoring device comprising: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute a power transformer state monitoring method in the above-mentioned embodiment 1. The power transformer state monitoring device in the embodiment of the present invention may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. The power transformer state monitoring device is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention. A power transformer condition monitoring device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM) 1004. In RAM 1004, various programs and data required for the operation of a power transformer condition monitoring device are also stored. The processing device 1001, ROM 1002, and RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow a power transformer condition monitoring device to communicate wirelessly or wired with other devices to exchange data. Although a power transformer condition monitoring device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0097] The present invention also provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, the steps of the above-mentioned method for monitoring the state of a power transformer are implemented. The computer program product provided by the present invention can solve a technical problem of monitoring the state of a power transformer. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as the beneficial effects of the method for monitoring the state of a power transformer provided by the above-mentioned embodiment, and will not be described in detail here.

[0098] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through a communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present invention are executed.

[0099] It should be understood that the various parts disclosed in the present invention can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0100] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for monitoring the state of a power transformer, characterized in that: The method comprises: Step S10: setting the sampling frequency and sampling duration of the transformer sound signal, the sampling frequency is 51200 Hz, the sampling duration is 2 seconds, and collecting the first sound signal of the transformer through the sound collection sensor, the first sound signal includes the low-frequency and high-frequency sound signals during the operation of the transformer; Step S20: performing windowing processing on the first sound signal, wherein the windowing processing uses a Hanning window to perform segmented processing on the sound signal, wherein the length of each signal segment is 12288 sampling points, and the overlapping length of adjacent signal segments is 4096 sampling points, and each signal segment is subjected to one-dimensional compression processing to generate a one-dimensional compressed signal sequence; Step S30: designing an adaptive cochlear filter bank, wherein the filter bank includes 18 filters, the center frequency of each filter matches the frequency spectrum distribution of the transformer sound signal, and the bandwidth range of the filter is 50 Hz to 1200 Hz. The one-dimensional compressed signal sequence is filtered by the filter bank to generate a cochlear characteristic coefficient matrix for each frequency band; Step S40: applying a multi-resolution analysis method of wavelet transform to the cochlear characteristic coefficient matrix, decomposing the one-dimensional compressed signal sequence into frequency sub-bands of different scales, extracting multi-resolution feature vectors by analyzing the features at different scales, and fusing them to generate an auditory feature parameter matrix; Step S50: Based on the mean and variance of the feature matrix, the auditory feature parameter matrix is ​​normalized to eliminate signal differences caused by different distances and intensities, thereby obtaining a normalized auditory feature matrix; Step S60: Obtain the standard auditory feature parameter matrix of the transformer under normal operating conditions, and calculate the information entropy weighted grey correlation between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, including weighting the feature parameters according to the entropy weights in the frequency domain and the time domain, and the formula is: Among them, Δ i represents the absolute difference between the standard auditory feature parameter matrix and the auditory feature parameter matrix to be tested in i features, Δ m1n Represents the minimum value of all feature difference values, Δ max represents the maximum value of all feature difference values, ρ is the information entropy weighted grey correlation degree between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, 0<ρ<1, E f It represents the entropy weight in the frequency domain, which represents the importance of the transformer sound signal at different frequency components. t It represents the entropy weight in the time domain, reflecting the importance of the features of the transformer sound signal in different time periods, and n is the total number of features involved in the correlation calculation; Step S70: If the value of the information entropy weighted grey correlation degree is greater than or equal to 0.8, the transformer operation state is determined to be normal; if it is less than 0.8, the transformer operation state is determined to be abnormal.

2. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In the step S10, the sound collection sensor is an electromagnetic sensor or a piezoelectric sensor, which can capture sound signals of different frequency ranges generated during the operation of the transformer.

3. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In step S10, the sampling frequency and sampling duration are adjusted according to the operating environment and type of different transformers.

4. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In the step S20, the window function used in the windowing process also includes a Gaussian window.

5. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In step S30, the design of the adaptive cochlear filter bank can dynamically adjust the bandwidth and center frequency of each filter according to the spectrum changes of the sound signal collected in real time.

6. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In the step S70, when the information entropy weighted grey correlation value is less than 0.8, it is determined that the transformer operation state is abnormal and then the sound and light alarm device is triggered.

7. The method for monitoring the state of a power transformer according to claim 1, characterized in that: In the step S50, the Z-score normalization method is used for normalization processing.

8. A power transformer status monitoring device, characterized in that: The device comprises: A signal sampling module is used to set the sampling frequency and sampling duration of the transformer sound signal, the sampling frequency is 51200 Hz, the sampling duration is 2 seconds, and the first sound signal of the transformer is collected by the sound collection sensor, the first sound signal includes the low-frequency and high-frequency sound signals during the operation of the transformer; A signal processing module, used to perform windowing processing on the first sound signal, wherein the windowing processing uses a Hanning window to perform segmented processing on the sound signal, the length of each signal segment is 12288 sampling points, the overlapping length of adjacent signal segments is 4096 sampling points, and each signal segment is subjected to one-dimensional compression processing to generate a one-dimensional compressed signal sequence; A filter design module is used to design an adaptive cochlear filter bank, wherein the filter bank includes 18 filters, the center frequency of each filter matches the frequency spectrum distribution of the transformer sound signal, and the bandwidth range of the filter is 50 Hz to 1200 Hz. The one-dimensional compressed signal sequence is filtered by the filter bank to generate a cochlear characteristic coefficient matrix for each frequency band; A parameter matrix generation module is used to apply a multi-resolution analysis method of wavelet transform to the cochlear characteristic coefficient matrix, decompose the one-dimensional compressed signal sequence into frequency sub-bands of different scales, extract multi-resolution feature vectors by analyzing the features at different scales, and fuse them to generate an auditory feature parameter matrix; A normalization processing module, used for normalizing the auditory feature parameter matrix based on the mean and variance of the feature matrix to eliminate signal differences caused by different distances and intensities, and obtaining a normalized auditory feature matrix; The information entropy weighted grey correlation calculation module is used to obtain the standard auditory feature parameter matrix of the transformer under normal operating conditions, and calculate the information entropy weighted grey correlation between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, including weighting the feature parameters according to the entropy weights in the frequency domain and time domain. The formula is: Among them, Δ i represents the absolute difference between the standard auditory feature parameter matrix and the auditory feature parameter matrix to be tested in i features, Δ m1n Represents the minimum value of all feature difference values, Δ max represents the maximum value of all feature difference values, ρ is the information entropy weighted grey correlation degree between the auditory feature parameter matrix of the transformer to be tested and the standard auditory feature parameter matrix, 0<ρ<1, E f It represents the entropy weight in the frequency domain, which represents the importance of the transformer sound signal at different frequency components. t It represents the entropy weight in the time domain, reflecting the importance of the features of the transformer sound signal in different time periods, and n is the total number of features involved in the correlation calculation; The fault judgment module is used to judge that if the value of the information entropy weighted grey correlation degree is greater than or equal to 0.8, the transformer operation state is judged to be normal; if it is less than 0.8, the transformer operation state is judged to be abnormal.

9. A power transformer condition monitoring device, characterized in that: The device comprises: A memory, a processor, and a power transformer state monitoring program stored in the memory and executable on the processor, wherein the power transformer state monitoring program implements the power transformer state monitoring method when executed by the processor.

10. A computer program product, characterized in that The computer program product comprises a power transformer state monitoring program, and the power transformer state monitoring program implements the power transformer state monitoring method when executed by a processor.

Images