Voiceprint-based transformer partial discharge detection method, apparatus, device and storage medium

Abstract

A voiceprint-based transformer partial discharge detection method, an apparatus, a device, and a storage medium. The method comprises: collecting a voiceprint signal in a transformer, and transmitting the collected voiceprint signal into a signal processing unit (220) for subsequent signal processing (S1); processing the voiceprint signal in the transformer by means of the signal processing unit (220) to obtain parameters (S2); determining the type of the transformer partial discharge by means of a partial discharge classifier unit (230) on the basis of the parameters (S3); and dividing the partial discharge state to obtain a detected state level classification result (S4).

Description

Transformer soundprint partial discharge detection method, device, equipment and storage medium

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on March 29, 2024, with application number 202410379824.8, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of voiceprint fault diagnosis and detection, for example, to a transformer voiceprint partial discharge detection method, device, equipment and storage medium. Background Art

[0003] Power transformers are crucial equipment in power systems, serving as their foundation and core. Failures in distribution transformers, in particular, can have a wide-ranging impact on a significant number of people. Economic development has led to a significant increase in both electrical equipment and total electricity consumption, placing higher demands on the operation of distribution transformers, requiring them to meet both stability and safety requirements. Partial discharge is the most common type of failure during transformer operation.

[0004] A common method for detecting partial discharge faults in transformers is to use oil chromatography to analyze the gas content of oil samples from the oil tank of a pole-mounted distribution transformer. Abnormalities in the oil sample must accumulate for a certain period of time before they can be detected through oil chromatography analysis. This accumulation period can take anywhere from six months to several years. Furthermore, taking transformer oil samples requires a power outage for the entire transformer, which defeats the goal of early detection, diagnosis, and treatment during transformer operation and is detrimental to stable operation. Therefore, early detection and no power outages are crucial indicators during transformer operation, and routine maintenance inspections should focus on this.

[0005] In recent years, fault diagnosis based on voiceprint features has gradually come into the view of transformer operation and maintenance personnel. By analyzing the voiceprint features of transformer operation faults, the goal of rapid diagnosis and early detection can be achieved without the need for more professional knowledge or power outages, which greatly improves the inspection efficiency of operation and maintenance personnel. Summary of the Invention

[0006] The present application provides a transformer soundprint partial discharge detection method, device, equipment and storage medium, which can improve the time of partial discharge fault diagnosis and achieve rapid diagnosis without power outage, thereby further improving the goal of early detection, early diagnosis and early treatment during transformer operation, and has significant significance for the stable and safe operation of the transformer.

[0007] This application provides a transformer soundprint partial discharge detection method, comprising:

[0008] Collect the voiceprint signal inside the transformer and transmit the collected voiceprint signal to the signal processing unit for subsequent signal processing;

[0009] The signal processing unit processes the original voiceprint signal inside the transformer to obtain parameters;

[0010] Determining the type of partial discharge of the transformer by a partial discharge classifier unit according to the parameters;

[0011] The partial discharge status is divided into categories to obtain the detection status level classification results.

[0012] In one or more embodiments, collecting the voiceprint signal inside the transformer includes: placing a sensor outside the transformer to collect the voiceprint signal inside the transformer.

[0013] In one or more embodiments, transmitting the collected voiceprint signal to the input signal processing unit includes: directly transmitting the collected voiceprint signal to the input signal processing unit through a network.

[0014] In one or more embodiments, the signal processing unit processes the original transformer internal voiceprint signal to obtain parameters, including:

[0015] After obtaining the original transformer internal voiceprint signal, the original transformer internal voiceprint signal is preprocessed according to the frequency band characteristics of the partial discharge voiceprint signal. The low-frequency signal is filtered out using a high-pass filter, and the signal containing the frequency band of the partial discharge voiceprint signal is retained. The formula for the high-pass filter operation is: y[n] = α(y[n-1] + x[n] - x[n-1])

[0016] Where y[n] is the current output of the original signal, x[n] is the current input of the original signal, x[n-1] is the previous input of the original signal, y[n-1] is the previous output of the original signal, and α is the coefficient that controls the cutoff frequency of the filter.

[0017] After obtaining the filtered signal, the partial discharge signal is very weak and needs to be processed. The Duffing oscillator can exhibit many key characteristics and can be used to extract effective information from weak signals. The formula of the Duffing oscillator is: x″+kx′-x+αx 3 =γcos(ω·t)

[0018] Where ω is the system driving force angular frequency, k is the damping ratio, and γ is the system built-in drive amplitude, which is usually 0. 3is the nonlinear term of the system. By controlling the coefficients k and α in the above formula, a weak signal can be obtained. However, solving nonlinear equations is often complex, especially with longer time series of the input signal. x″ and x′ are the second-order derivative and first-order derivative, representing acceleration and velocity, respectively, and t is time.

[0019] After obtaining the preprocessed signal, time-synchronized phase alignment is performed on a long sequence of signals, so that the discrete signals of a long sequence are compressed into a space with a specified time phase width, so that the discrete signals are superimposed and the distribution of the discrete signals is aggregated. The formula for time-synchronized phase alignment is:

[0020] Where f(s) is the output of the signal, s is the unit signal strength, t is the time of the signal, T is the specified time, and n is the maximum integer period of T.

[0021] After completing the conversion of long time series signals to short time series signals, the short time series signals are divided into two parts, the left time series S1 and the right time series S2, and the signal strengths of the initial points on the left and right sides are extracted respectively, s0,s mid ,s end (i.e. the starting point, middle point and end point of the left time series S1 and the right time series S2), the time coordinates of the three points (t0, s0) are obtained from the time points corresponding to the three values. mid ,s mid ),(t end ,s end ), through the characteristics of the linear function, the three coordinate points are fitted into two linear curves. The formula for fitting the linear curve is:

[0022] Among them, x1, y1 are the curves fitted by the left time series, k1 is the slope of the left time series curve, x2, y2 are the curves fitted by the right time series, and k2 is the slope of the right time series curve;

[0023] Arrange the discrete signal points in the left time series and the right time series in ascending order to obtain two ordered discrete sequences S1', {s0', s1', s2', ... s (t-2) / 2 ',s (t-1) / 2 '} and S2', {s t / 2 ',s (t+1) / 2 ',s (t+2) / 2 '……s t-1 ',s t'}; Calculate the minimum value, the value at 1 / 4, the value at 1 / 2, the value at 3 / 4 and the maximum value of the ordered discrete signal points in the left time series, and obtain the time coordinates of the five points (t0', s0'), (t (t-1) / 4 ',s (t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '),(t 3(t-1) / 4 ',s 3(t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '), calculate the average distance z1 of the five points to y1 and compare it with the distance threshold ε; similarly, take five points from the right time series, use the five-point calculation method to calculate the average distance z2 of the five points to y2, and compare it with the distance threshold ε;

[0024] After obtaining the short time series, the short time series signal is divided into two parts, namely the front time series L1 and the back time series L2, and the signal intensity ratio of the front time series and the back time series is calculated respectively. The calculation formula of the signal intensity ratio is:

[0025] Among them, p is the signal intensity ratio, l is the unit signal intensity, l average is the average value of the whole signal; thus a new time series P1 and P2 are formed, and the three values ​​of the initial moment, the truncation point of the time series and the last moment are extracted respectively, p0, p mid ,p end , the three coordinate points are composed of the time points corresponding to the three values, (t0, p0), (t mid ,p mid ),(t end ,p end ), through the characteristics of the linear function, the three coordinate points are fitted into two linear curves; the formula for fitting the linear curve is:

[0026] Among them, x3, y3 are the curves fitted by the new front time series, k3 is the slope of the new front time series curve, x4, y4 are the curves fitted by the new back time series, and k4 is the slope of the new back time series curve;

[0027] Arrange the discrete signal points in the new before and after time series from small to large to obtain two ordered discrete sequences L1', {l0',l1',l2',...l m-1 ',l m '} and L2', {l t-m ',l t-m-1 ',l t-m-2 ',……l t-1 ',l t'}, use the 5-point calculation method to calculate the coordinate points of the 5 values ​​in the L1' and L2' sequences; calculate the average values ​​z3, z4 of the 5 values ​​in the L1' and L2' sequences and compare them with the distance threshold ε.

[0028] In one or more embodiments, the parameters include: the calculated left time series curve y1, the left time series average distance z1, the right time series curve y2, the right time series average distance z2, the front time series curve y3, the front time series average distance z3, the rear time series y4, the rear time series average distance z4 and the distance threshold ε.

[0029] In one or more embodiments, determining the type of transformer partial discharge by the partial discharge classifier unit includes:

[0030] Compare the four average distance values ​​z1, z2, z3, and z4 with the distance threshold ε and make a judgment;

[0031] If z1 ≥ ε, set the flag to 1, otherwise set it to 0; similarly, if z2, z3, z4 are greater than or equal to ε, set the flag to 1, otherwise set it to 0;

[0032] The four flag bits are accumulated to obtain the partial discharge classification result.

[0033] In one or more embodiments, the partial discharge status is divided by the accumulated values ​​of the four flag bits.

[0034] In one or more embodiments, during the classification of the partial discharge status, normal indicates that there is no partial discharge signal; concern indicates that there is a slight partial discharge and attention is required, and the condition of the equipment needs to be continuously monitored to prevent equipment deterioration; and alarm indicates that a partial discharge signal is generated and is relatively obvious, requiring operation and maintenance personnel to investigate the situation.

[0035] This application also provides a transformer soundprint partial discharge detection device, comprising:

[0036] a signal acquisition unit configured to acquire a voiceprint signal inside the transformer and transmit the acquired voiceprint signal to an input signal processing unit for subsequent signal processing;

[0037] a signal processing unit configured to process the voiceprint signal inside the transformer to obtain parameters;

[0038] The partial discharge classifier unit is configured to update the state of the flag bit by comparing the distance threshold with the calculated distance value;

[0039] The analysis and processing unit is configured to classify the status of the partial discharge and obtain a detection status level classification result.

[0040] In one or more embodiments, the device is used to perform the transformer soundprint partial discharge detection method as described in any one of the above items.

[0041] The present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, a transformer voiceprint partial discharge detection method is implemented.

[0042] The present application also provides a computer-readable storage medium having a computer program stored thereon, which implements a transformer soundprint partial discharge detection method when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] FIG1 is a flow chart of a transformer soundprint partial discharge detection method using an impulse phase method;

[0044] FIG2 is a flow chart of a method for calculating an impulse phase in an embodiment of the present application;

[0045] FIG3 is a waveform diagram of the original signal collected in an embodiment of the present application;

[0046] FIG4 is a time-compressed signal diagram according to an embodiment of the present application;

[0047] FIG5 is a diagram showing the results of the test in the embodiment of the present application;

[0048] FIG6 is a schematic structural diagram of a transformer soundprint partial discharge detection device according to an embodiment of the present application;

[0049] FIG7 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] All features disclosed in all embodiments in this specification, or steps in all methods or processes implicitly disclosed, except for mutually exclusive features and / or steps, can be combined and / or expanded or replaced in any manner.

[0051] Fault diagnosis based on voiceprint features is gradually coming into the view of transformer operation and maintenance personnel. By analyzing the voiceprint features of transformer operation faults, the goal of rapid diagnosis and early detection can be achieved without the need for more professional knowledge or power outages, which greatly improves the inspection efficiency of operation and maintenance personnel.

[0052] As shown in Figures 1 to 5, this application provides a transformer soundprint partial discharge detection method using the impulse phase method, providing a diagnostic method for partial discharge during transformer operating faults. The following describes this method in conjunction with an embodiment. The method includes the following steps.

[0053] S1: Collect the voiceprint signal inside the transformer and transmit the collected voiceprint signal to the signal processing unit for subsequent signal processing.

[0054] The collecting of the voiceprint signal inside the transformer includes placing a sensor outside the transformer to collect the voiceprint signal inside the transformer.

[0055] The transmitting the collected voiceprint signal to the signal processing unit includes: directly transmitting the collected voiceprint signal to the signal processing unit through a network.

[0056] S2: The original transformer internal voiceprint signal is processed by the signal processing unit to obtain parameters. The operation steps are as follows:

[0057] S21: After obtaining the original voiceprint signal through S1, pre-process the original voiceprint signal according to the frequency band characteristics of the partial discharge voiceprint signal, and use high-pass filtering to filter out the low-frequency signal and only retain the signal containing the frequency band of the partial discharge voiceprint signal. The formula of the high-pass filtering operation is: y[n] = α(y[n-1] + x[n] - x[n-1])

[0058] Where y[n] is the current output of the original signal, x[n] is the current input of the original signal, x[n-1] is the previous input of the original signal, y[n-1] is the previous output of the original signal, and α is the coefficient that controls the cutoff frequency of the filter.

[0059] The occurrence frequency of partial discharge is generally higher than 10 kHz. In this embodiment, the signal with a frequency lower than 10 kHz is subjected to high-pass filtering.

[0060] S211: After obtaining the filtered signal in step S21, the partial discharge signal is very weak and needs to be processed. The Duffing oscillator can exhibit many key characteristics and can be used to extract effective information from weak signals. The formula of the Duffing oscillator is: x″+kx′-x+αx 3 =γcos(ω·t)

[0061] Where ω is the system driving force angular frequency, k is the damping ratio, and γ is the system built-in drive amplitude, which is usually 0. 3 It is the nonlinear term of the system. By controlling the k and α coefficients in the above formula, a weak signal can be obtained. However, solving nonlinear equations is usually more complicated, especially when the time series of the input signal is longer, the complexity will be higher.

[0062] S22: After obtaining the pre-processed signal in step S211, perform time synchronization phase alignment on the signal of a long time series, so that the discrete signal of a long sequence is compressed into a space with a specified phase width at a specific time, so that the discrete signals are superimposed and the distribution of the discrete signals is concentrated, which is more conducive to the analysis and processing of subsequent operations. The formula for time synchronization phase alignment is:

[0063] Where f(s) is the output of the signal, s is the unit signal strength, t is the time of the signal, T is the specified time, and n is the maximum integer period of T.

[0064] Assuming that the power frequency signal is 50 Hz and the period is 0.02 seconds, in this embodiment, the long time series signal is compressed to within 0.02 seconds, and time compression is performed every 0.02 seconds to unify the time of each segment into the power frequency period.

[0065] S23: Step S22 completes the conversion of long time series to short time series, and divides the short time series into two parts, namely the left time series S1 and the right time series S2, and extracts the signal strength of the initial points on the left and right sides, s0, s mid ,s end , and then the time coordinates of the three points (t0, s0) are obtained from the time points corresponding to the three values. mid ,s mid ),(t end ,s end ), through the characteristics of linear function, three coordinate points can be fitted into two linear curves. The formula for fitting linear curve is:

[0066] Where x1 and y1 are the curves fitted to the left time series, and k1 is the slope of the left time series curve. x2 and y2 are the curves fitted to the right time series, and k2 is the slope of the right time series curve. In this embodiment, the left and right time series are half the power frequency cycle time, that is, 0.01 seconds. 0 to 0.01 seconds is the left time series, and 0.01 to 0.02 seconds is the right time series.

[0067] S24: Arrange the discrete signal points in the left and right time series in ascending order to obtain two ordered discrete sequences S1', {s0', s1', s2', ... s (t-2) / 2 ',s (t-1) / 2 '} and S2', {s t / 2 ',s (t+1) / 2 ',s (t+2) / 2 '……s t-1 ',s t'}, in order to eliminate the mutation of sequence data, this application adopts the 5-point calculation method. Calculate the minimum value, the value at 1 / 4, the value at 1 / 2, the value at 3 / 4 and the maximum value of the ordered discrete signal points in the left time series. The time coordinates of these 5 points are (t0', s0'), (t (t-1) / 4 ',s (t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '),(t 3(t-1) / 4 ',s 3(t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '), calculate the average distance z1 of these five points to y1 and compare it with the distance threshold ε; similarly, take five points from the right time series, calculate the average distance z2 of these five points to y2, and compare it with the distance threshold ε. In this embodiment, the distance threshold ε is 2.

[0068] S25: The short time series obtained in step S22 is divided into two parts. Different from the left and right time series in step S23, the two time series in this step are the front time series L1 and the back time series L2. In this embodiment, the front and back time series take one-fourth of the power frequency cycle time, that is, 0.025 seconds. 0 to 0.0025 seconds is the left time series, and 0.0025 seconds to 0.02 seconds is the right time series. The signal strength ratio of the front and back time series is calculated respectively. The calculation formula of the signal strength ratio is:

[0069] Among them, p is the signal intensity ratio, l is the unit signal intensity, l average is the average value of the whole signal. Thus, a new time series P1 and P2 are formed, and the three values ​​of the initial moment, the truncation point of the time series and the last moment are extracted respectively, p0, p mid ,p end , and then the three coordinate points are formed by the time points corresponding to the three values, (t0, p0), (t mid ,p mid ),(t end ,p end ), through the characteristics of linear function, three coordinate points can be fitted into two linear curves. The formula for fitting linear curve is:

[0070] Among them, x3, y3 are the curves fitted by the new front time series, k3 is the slope of the new front time series curve, x4, y4 are the curves fitted by the new back time series, and k2 is the slope of the new back time series curve.

[0071] S26: Arrange the discrete signal points in the new before and after time series from small to large to obtain two ordered discrete sequences L1', {l0',l1',l2',...l m-1 ',l m '} and L2', {l t-m ',l t-m-1 ',l t-m-2 ',……l t-1 ',l t '}, calculate the coordinate points of the five values ​​in the L1' and L2' sequences, and the calculation method of the five values ​​is consistent with step S24. Calculate the average values ​​z3 and z4 of the five values ​​in the L1' and L2' sequences and compare them with the distance threshold ε.

[0072] S3: In step S2, the left time series curve y1, the left time series average distance z1, the right time series curve y2, the right time series average distance z2, the front time series curve y3, the front time series average distance z3, the back time series y4, the back time series average distance z4, and the distance threshold ε are calculated. The above parameters are used to determine the type of partial discharge of the transformer through the partial discharge classifier unit. The steps are as follows:

[0073] In step S3, the state of the partial discharge (eg, normal, concern, and alarm) is obtained. S31: The four average distance values ​​z1, z2, z3, and z4 are compared with the distance threshold ε respectively to make the next judgment.

[0074] S32: If z1 ≥ ε, set the flag bit to 1; otherwise, set the flag bit to 0. Similarly, if z2, z3, or z4 are greater than or equal to ε, set the flag bit to 1; otherwise, set the flag bit to 0. That is, if one of z2, z3, or z4 is greater than or equal to ε, set the flag bit of the corresponding one to 1. For example, if z2 is greater than or equal to ε, set the flag bit corresponding to z2 to 1.

[0075] S33: Accumulate the four flag bits to obtain a classification result of the partial discharge.

[0076] As shown in Table 1, the statistical results of the four flag bits are

[0077] Table 1

[0078] S4: The accumulated flag values ​​described in S33 are used to classify the PD status. Normal indicates no PD signal; Concern indicates a mild PD signal that requires continued attention and prevents deterioration; and Alarm indicates a significant PD signal that requires investigation in conjunction with O&M. Table 2 shows the correspondence between the accumulated flag values ​​and the status levels. In S4, the PD status is classified into three states: Normal, Concern, and Alarm.

[0079] Table 2

[0080] In this embodiment, the cumulative value of the flag bit obtained from Table 1 is 4, which corresponds to the alarm level according to the level status in Table 2. The suggestion given is that the operation and maintenance personnel need to check the fault point and perform fault elimination operations at the fault point.

[0081] S5: In steps S2 to S4, a time series of partial discharge detection results is obtained. However, the results of a time series are often not robust. The results of a time series are then processed using a support vector machine (SVM) classification process. The steps are as follows:

[0082] S51: The three detection results obtained in each time series, normal, concern, and warning, are used as inputs to the classifier. Two of them are grouped together to form three key pairs: (normal, concern), (normal, warning), and (concern, warning).

[0083] S52: Define a hyperplane to separate each category, the formula is: f(x) = ω·x+b

[0084] Where ω is the normal vector and b is the bias.

[0085] S53: The farther the sample points on both sides of the hyperplane are, the better the classification effect will be. The equation of the plane is adjusted by continuously updating the values ​​of ω and b.

[0086] S54: Find ω and b to maximize the interval while ensuring that all sample points are correctly classified. And satisfy y i (ω·x i +b)≥1, then the two parameters at this moment are the optimal parameters.

[0087] S55: Following the above steps, the three key pairs are classified and reorganized in sequence to obtain the final result. The resulting suggestion is that the operation and maintenance personnel need to continue troubleshooting the fault point and eliminate the fault at the fault point.

[0088] The transformer soundprint partial discharge detection method provided in the present application can determine partial discharge and can achieve real-time detection, real-time detection conclusions, and online detection and processing. This greatly improves the time for transformer partial discharge detection, eliminates the need for power outages, and improves the efficiency of daily inspections by operation and maintenance personnel. In addition, in addition to improving detection time, the method of the present application improves detection accuracy and can detect weak partial discharge signals, which is something that traditional oil chromatography detection cannot do, and live detection is also something that traditional oil chromatography cannot do.

[0089] The present application also provides a transformer soundprint partial discharge detection device. FIG6 is a schematic diagram of the structure of a transformer soundprint partial discharge detection device provided in the present application. As shown in FIG6, the device includes:

[0090] The signal collection unit 210 is configured to collect original voiceprint signals;

[0091] The signal processing unit 220 is configured to perform preprocessing, time compression, linear curve fitting and distance calculation on the signal, thereby obtaining the distance from the point to the fitting curve and comparing it with the size of the distance threshold;

[0092] The partial discharge classifier unit 230 is configured to update the state of the flag bit based on the difference between the distance threshold and the calculated distance value;

[0093] The analysis and processing unit 240 is configured to calculate the cumulative value of the flag bit and provide a detection status level classification result.

[0094] a signal acquisition unit configured to acquire a voiceprint signal inside the transformer and transmit the acquired voiceprint signal to an input signal processing unit for subsequent signal processing;

[0095] a signal processing unit configured to process the voiceprint signal inside the transformer to obtain parameters;

[0096] a partial discharge classifier unit configured to obtain a partial discharge state according to the parameters;

[0097] The analysis and processing unit is configured to classify the status of the partial discharge and obtain a detection status level classification result.

[0098] The transformer soundprint partial discharge detection device provided in the embodiment of the present application can execute the transformer soundprint partial discharge detection method provided in any embodiment of the present application, and has the corresponding functional modules and effects of the execution method.

[0099] An embodiment of the present application also provides an electronic device. FIG7 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0100] As shown in FIG7 , the electronic device 10 includes at least one processor 11 and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12 and a random access memory (RAM) 13. The memory stores a computer program that can be executed by the at least one processor 11. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the ROM 12 or the computer program loaded from the storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0101] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0102] Processor 11 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of processor 11 include a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any other suitable processor, controller, microcontroller, etc. Processor 11 executes the various methods and processes described above, such as the transformer voiceprint partial discharge detection method.

[0103] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a transformer voiceprint partial discharge detection method.

[0104] In some embodiments, the transformer soundprint partial discharge detection method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the transformer soundprint partial discharge detection method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to execute the transformer soundprint partial discharge detection method via any other appropriate means (e.g., via firmware).

[0105] In the context of the present application, computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage medium can include electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage medium can be a machine-readable signal medium. Machine-readable storage medium includes an electrical connection based on one or more lines, a portable computer disk, a hard disk, RAM, ROM, erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device or any suitable combination of the foregoing. Storage medium can be a non-transitory storage medium.

[0106] The units involved in the embodiments described in this application can be implemented by software or hardware, and the units described can also be set in a processor. In particular, the names of these units do not constitute limitations on the units themselves.

[0107] According to one aspect of an embodiment of the present application, a computer program product or computer program is provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the transformer soundprint partial discharge detection method provided in the various optional implementations described above.

Claims

1. A transformer soundprint partial discharge detection method, comprising: Collect the voiceprint signal inside the transformer and transmit the collected voiceprint signal to the signal processing unit for subsequent signal processing; Processing the voiceprint signal inside the transformer by the signal processing unit to obtain parameters; Obtaining a partial discharge state through a partial discharge classifier unit according to the parameters; The partial discharge status is divided into categories to obtain the detection status level classification results.

2. The method according to claim 1, wherein The collecting of the voiceprint signal inside the transformer includes: placing a sensor outside the transformer to collect the voiceprint signal inside the transformer.

3. The method according to claim 1, wherein The transmitting the collected voiceprint signal to the signal processing unit includes: directly transmitting the collected voiceprint signal to the signal processing unit through a network.

4. The method according to claim 1, wherein The processing of the voiceprint signal inside the transformer by the signal processing unit to obtain parameters includes: After obtaining the voiceprint signal inside the transformer, the voiceprint signal inside the transformer is preprocessed according to the frequency band characteristics of the partial discharge voiceprint signal, wherein the preprocessing includes: using high-pass filtering to filter the low-frequency signal and retain the signal containing the frequency band of the partial discharge voiceprint signal; wherein the formula of the high-pass filtering operation is: y[n]=α(y[n-1]+x[n]-x[n-1]) Where y[n] is the current output of the voiceprint signal inside the transformer, x[n] is the current input of the voiceprint signal inside the transformer, x[n-1] is the previous input of the voiceprint signal inside the transformer, y[n-1] is the previous output of the voiceprint signal inside the transformer, and α is the coefficient that controls the cutoff frequency of the filter; After obtaining the preprocessed signal, a time-synchronous phase alignment is performed on a long-time series of discrete signals, so that the long-time series of discrete signals is compressed into a space with a selected time-specified phase width to obtain a short-time series signal, so that the discrete signals are superimposed and the distribution of the discrete signals is aggregated; wherein the formula for the time-synchronous phase alignment is: Where f(s) is the output of the signal, s is the unit signal strength, t is the time of the signal, T is the specified time, and n is the maximum integer period of T. After completing the conversion of the long time series discrete signal to the short time series signal, the short time series signal is divided into two parts, namely the left time series S1 and the right time series S2, and the signal strengths of the initial points on the left and right sides and the truncation points of the left time series S1 and the right time series S2 are extracted respectively, s0,s mid ,s end , the time coordinates of the three points (t0, s0) are obtained from the time points corresponding to the three values. mid ,s mid ),(t end ,s end ), through the characteristics of the linear function, the time coordinate points of the three points are fitted into two linear curves. The formula for fitting the linear curve is: Wherein, x1, y1 are the curves fitted by the left time series, k1 is the slope of the curve fitted by the left time series, x2, y2 are the curves fitted by the right time series, k2 is the slope of the curve fitted by the right time series; Arrange the discrete signal points in the left time series and the right time series in ascending order to obtain two ordered discrete sequences S1', {s0', s1', s2', ... s (t-2) / 2 ',s (t-1) / 2 '} and S2', {s t / 2 ',s (t+1) / 2 ',s (t+2) / 2 '……s t-1 ',s t '}; Calculate the minimum value, the value at 1 / 4, the value at 1 / 2, the value at 3 / 4 and the maximum value of the ordered discrete signal points in the left time series, and obtain the time coordinates of the five points (t0', s0'), (t (t-1) / 4 ',s (t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '),(t 3(t-1) / 4 ',s 3(t-1) / 4 '),(t (t-1) / 2 ',s (t-1) / 2 '), calculate the average value of the distances from the five points to y1: the left time series average distance z1, and compare z1 with the distance threshold ε; take five points in the right time series, which are the minimum value, the value at 1 / 4, the value at 1 / 2, the value at 3 / 4 and the maximum value of the right time series, and use the five-point calculation method to calculate the average value of the distances from the five points to y2: the right time series average distance z2, and compare z2 with the distance threshold ε; The short time series signal is divided into two parts, namely the front time series L1 and the back time series L2, and the signal strength ratio of the front time series and the back time series is calculated respectively. The calculation formula of the signal strength ratio is: Among them, p is the signal intensity ratio, l is the unit signal intensity, l average is the average value of the whole signal; thus a new front time series P1 and a new back time series P2 are formed, and the three values ​​of the initial moment, the truncation point of the front and back time series and the last moment are extracted respectively, p0, p mid ,p end , the three coordinate points are composed of the time points corresponding to the three values, (t0, p0), (t mid ,p mid ),(t end ,p end ), through the characteristics of the linear function, the three coordinate points are fitted into two linear curves; the formula for fitting the linear curve is: Wherein, x3, y3 are the curves fitted by the new front time series, k3 is the slope of the new front time series curve, x4, y4 are the curves fitted by the new back time series, k4 is the slope of the new back time series curve; Arrange the discrete signal points in the new front time series and the new back time series from small to large to obtain two ordered discrete sequences L1', {l0',l1',l2',...l m-1 ',l m '} and L2', {l t-m ',l t-m-1 ',l t-m-2 ',……l t-1 ',l t '}, use the 5-point calculation method to calculate the coordinate points of the 5 values ​​in the L1' and L2' sequences; calculate the average values ​​of the 5 values ​​in the L1' and L2' sequences, which are: the average distance z3 of the previous time series and the average distance z4 of the subsequent time series, and compare z3 and z4 with the distance threshold ε respectively.

5. The method according to claim 4, wherein The parameters include: the calculated left time series curve y1, the left time series average distance z1, the right time series curve y2, the right time series average distance z2, the front time series curve y3, the front time series average distance z3, the back time series y4, the back time series average distance z4 and the distance threshold ε.

6. The method according to claim 5, wherein: The obtaining of the partial discharge state by a partial discharge classifier unit according to the parameters includes: Compare the four average distance values ​​z1, z2, z3, and z4 with the distance threshold ε respectively and make a judgment; In response to z1 being greater than or equal to ε, the flag corresponding to z1 is set to 1; in response to z1 being less than ε, the flag corresponding to z1 is set to 0; in response to one of z2, z3, and z4 being greater than or equal to ε, the flags corresponding to z2, z3, and z4 are set to 1; in response to at least one of z2, z3, and z4 being less than ε, the flags corresponding to z2, z3, and z4 are set to 0, resulting in four flags; The four flag bits are accumulated to obtain a classification result of the partial discharge.

7. The method according to claim 6, wherein: The dividing the partial discharge state includes dividing the partial discharge state according to the accumulated value of the four flag bits.

8. The method according to claim 7, wherein: In the process of classifying the partial discharge status, normal is used to indicate that there is no partial discharge signal; concern is used to indicate that there is a slight partial discharge; and alarm is used to indicate that a partial discharge signal is generated.

9. A transformer soundprint partial discharge detection device, comprising: a signal acquisition unit configured to acquire a voiceprint signal inside the transformer and transmit the acquired voiceprint signal to an input signal processing unit for subsequent signal processing; a signal processing unit configured to process the voiceprint signal inside the transformer to obtain parameters; a partial discharge classifier unit configured to obtain a partial discharge state according to the parameters; The analysis and processing unit is configured to classify the status of the partial discharge and obtain a detection status level classification result.

10. An electronic device comprising a memory, at least one processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the transformer soundprint partial discharge detection method according to any one of claims 1 to 8 is implemented.

11. A computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the transformer soundprint partial discharge detection method according to any one of claims 1 to 8 is implemented.

Images