Abnormal sound recognition and detection device for high-voltage equipment of transformer substation

Abstract

The invention relates to the technical field of transformer substation high-voltage equipment detection, and discloses a transformer substation high-voltage equipment abnormal sound recognition and detection device which comprises a voiceprint recognition unit, an ozone gas detection unit, a partial discharge sensing unit, a data processing and analysis unit and an early warning unit. The voiceprint recognition unit collects the operation sound of the equipment in real time; the ozone gas detection unit detects the ozone concentration in the air; the partial discharge sensing unit collects a partial discharge signal of the equipment; and the data processing and analyzing unit is used for processing and analyzing based on the standby operation sound data, the ozone concentration data and the partial discharge signal of the equipment, and carrying out accurate identification and early warning on the abnormal state of the equipment based on an analysis result. The device has the advantages of multi-dimensional monitoring, real-time performance, visualization, intelligence and the like, and can realize effective monitoring and fault early warning of the state of the high-voltage equipment; actual tests show that the device can significantly improve the operation efficiency and safety of the power grid and reduce the operation and maintenance cost.

Description

Technical Field

[0001] The present invention relates to the technical field of substation high-voltage equipment detection, and in particular to a device for identifying and detecting abnormal sounds of substation high-voltage equipment. Background Art

[0002] With the continuous growth of electricity demand, the requirements for safe operation and reliability of substations are becoming higher and higher. Regular inspection and maintenance, as an important part of substation management, can timely discover and deal with potential safety hazards, prevent the occurrence of faults, and ensure the stable operation of the power system. However, when substation operation and maintenance personnel enter various high-voltage places in the substation, such as: high-voltage equipment area, GIS room, high-voltage distribution room and other places, they often hear abnormal sounds of electrical equipment or smell pungent ozone. The current detection instruments cannot distinguish whether the abnormal phenomenon of the equipment is caused by electrical quantity faults or non-electrical quantity faults. When operation and maintenance personnel approach the faulty equipment for inspection and verification, there is a risk of personal safety. They can only judge the nature of the abnormal phenomenon of the equipment and the fault point based on experience, which often leads to two consequences. First, inaccurate judgment causes the expansion of the scope of equipment power outage, affecting the reliability of power supply; second, the nature of the equipment failure cannot be effectively judged, resulting in long-term "disease" operation of the equipment, which ultimately causes equipment damage and even endangers the life safety of the operation and maintenance personnel.

[0003] With the rapid development of the power industry, the operation status monitoring and fault diagnosis technology of high-voltage equipment in substations has gradually become a research hotspot. The traditional manual inspection method is not only inefficient, but also endangers the personal safety of operation and maintenance personnel and is difficult to detect potential equipment hazards. Therefore, the development of a device that can monitor and accurately identify abnormal sounds of high-voltage equipment in real time has important practical significance and application value. Summary of the invention

[0004] In response to the above problems, the present invention proposes a device for identifying and detecting abnormal sound of high-voltage equipment in a substation. The device integrates multiple technologies such as sound recognition, ozone recognition and partial discharge test data recognition, realizes multi-dimensional monitoring of the status of high-voltage equipment, and improves the accuracy and reliability of fault identification.

[0005] The technical solution adopted by the present invention is:

[0006] A device for identifying and detecting abnormal sound of high-voltage equipment in a substation, comprising a voiceprint recognition unit, an ozone gas detection unit, a partial discharge sensor unit, a data processing and analysis unit and an early warning unit; the voiceprint recognition unit, the ozone gas detection unit and the partial discharge sensor unit are all connected to the data processing and analysis unit for data transmission; the voiceprint recognition unit collects the operating sound of the equipment in real time; the ozone gas detection unit detects the ozone concentration in the air; the partial discharge sensor unit collects the partial discharge signal of the equipment; the data processing and analysis unit processes and analyzes the operating sound data, ozone concentration data and partial discharge signal of the equipment, the data processing and analysis unit is controlled and connected to the early warning unit, and accurately identifies and warns the abnormal state of the equipment based on the analysis result.

[0007] Furthermore, the voiceprint sensing unit includes four parts: a sound collection module, a sound processing module, a feature extraction module and a pattern recognition module. The sound collection module is connected to the sound processing module for data transmission, the sound processing module is connected to the feature extraction module for data transmission, and the feature extraction module is connected to the pattern recognition module for data transmission.

[0008] Furthermore, the ozone gas detection unit adopts an ozone sensor.

[0009] Furthermore, the partial discharge sensing unit adopts an air-type ultrasonic partial discharge sensor.

[0010] Furthermore, the abnormal sound recognition and detection device for high-voltage equipment in a substation includes a detector host, and the data processing and analysis unit, the early warning unit, and the sound processing module, feature extraction module and pattern recognition module of the voiceprint recognition unit are all installed in the detector host; the data processing and analysis unit is a working machine with a microprocessor; the early warning unit is an audible and visual alarm; the sound collection module, ozone gas detection unit, partial discharge sensor unit and other corresponding sensors of the voiceprint recognition unit communicate with the detector host wirelessly; the detector host has a built-in lithium battery to power the data processing and analysis unit, the early warning unit, the sound processing module, the feature extraction module and the pattern recognition module; the detector host is equipped with a USB main mode interface, a 2.4G wireless communication interface and a backup interface.

[0011] Furthermore, the voiceprint recognition unit performs spectrum analysis and time domain analysis on the sound signal as follows:

[0012] Time domain analysis: the waveform of the sound, by observing whether there are regular signals in the waveform, combined with subsequent spectrum and time-spectrum analysis;

[0013] Spectral analysis: It converts the time domain signal into frequency domain representation and analyzes its frequency components. This conversion is achieved through fast Fourier transform FFT. The specific Fourier transform formula is:

[0014] F(ω )=∫f(t)e-iwtdt

[0015] Where F(ω) is the Fourier transform result; f(t) is the original signal or time domain function; e -iwt It is a complex function related to frequency w and time t. It is used as a basis function in Fourier transform to represent the basic components of the frequency domain signal. i is an imaginary unit. dt represents the differential of the integral variable, and t is the integral variable.

[0016] Time-frequency spectrum analysis: Use wavelet transform to obtain the time-frequency characteristics of the signal. The formula is:

[0017]

[0018] In the formula, WT(a,τ) represents the result of wavelet transform; f(t) is the original signal, which refers to the signal in the time domain; ψ is the wavelet function, which is used to analyze the local characteristics of the signal; a is the scale parameter, which is used to control the expansion degree of the wavelet; τ is the translation parameter, which is used to control the movement of the wavelet function on the time axis; t is the integral variable; is the regularization factor, which is used to adjust the amplitude of wavelet transform;

[0019] Discrete Fourier Transform DFT: DFT is an important tool for obtaining the frequency domain characteristics of signals. Its calculation formula is:

[0020]

[0021] Where X(k) represents the complex value corresponding to the frequency index k of the discrete frequency domain signal after DFT; x(n) represents the sample value corresponding to the time index n of the discrete signal in the time domain; N is the length of the signal, that is, the total number of samples in the signal x(n); n is the sample index of the time domain signal; k refers to the index in the frequency domain; W N is the root of unity of a complex number, used as a basis function in DFT calculations; j represents the imaginary unit;

[0022] The computational complexity of DFT: According to the given DFT calculation formula, each calculation of a frequency point X(k) requires N complex multiplications and N-1 complex additions; calculating X(k) of N points requires N2 complex multiplications and N*(N-1) complex additions; when x(n) is a real number, calculating the DFT of N points requires 2*N2 real multiplications and 2*N*(N-1) real additions;

[0023] Decomposition process of discrete Fourier transform: by decomposing the original signal into odd and even values, and using the correlation property of WN, an N-point DFT is transformed into a combination of two N / 2-point DFTs;

[0024] The waveform after Fourier transformation is then transformed through wavelet transformation to finally form a spectrum, thereby judging the operating status of the equipment and whether there are any abnormalities.

[0025] Furthermore, the sound collection module uses a highly sensitive microphone array as a voiceprint sensor to capture the weak sounds generated during the operation of the device;

[0026] The sound processing module pre-processes the collected sound signals, including: removing environmental noise and background noise to improve the signal-to-noise ratio of the sound signal; filtering out high-frequency and low-frequency noise to retain useful sound features; and improving the clarity and recognition of the sound signal;

[0027] The feature extraction module extracts feature parameters reflecting the device status from the processed sound signal to identify different types of noise; by analyzing the spectral characteristics of the noise, different noise sources can be distinguished;

[0028] The pattern recognition module presents the results from the processed data in the form of a graph, uses a machine learning algorithm to learn and classify the extracted feature parameters, and builds a model to automatically learn the relationship between sound features and device status, thereby identifying abnormal device sounds.

[0029] Furthermore, under abnormal circumstances, the sound characteristics will change; the pattern recognition module detects abnormalities in a timely manner by comparing the voiceprints in normal and abnormal states.

[0030] Furthermore, the circuit of the sound processing module includes a filter circuit, and the filter circuit includes a second-order active low-pass filter and a second-order high-pass filter;

[0031] The transfer function of a second-order low-pass filter is:

[0032]

[0033] Where G(s) is the filter transfer function; U 0 (s) is the Laplace transform of the output signal; U i (s) is the Laplace transform of the input signal; G 0 Center frequency gain, which indicates the gain of the filter at the center frequency; ω m is the natural angular frequency, which indicates the natural frequency of the system; ξ is the damping coefficient, which determines the response characteristics of the system; S is the complex frequency variable, which indicates the frequency domain characteristics of the signal;

[0034] Using the frequency normalization method, the transfer function of the above second-order low-pass filter is:

[0035]

[0036] Where λ is the normalized frequency parameter, which is usually related to the characteristic frequency of the system; G m is the normalized gain, which is usually set to 1 during the normalization process;

[0037] The model of the second-order high-pass filter includes two RC filter circuits and a common-mode proportional amplifier circuit. It has high input impedance and low output impedance. Signals above the cutoff frequency pass through, and signals below the cutoff frequency are attenuated 100 times for every 10 times higher frequency.

[0038] The second-order high-pass filter has a bandpass gain of:

[0039]

[0040] In the formula, A γp is the passband gain; R f is the feedback resistor; R 1 is the input resistance;

[0041] The transfer function of a second-order high-pass filter is:

[0042]

[0043] In the formula, A γ (s) represents the frequency response transfer function; s is the complex frequency variable; C is the capacitance value; R is the resistance value;

[0044] The frequency response of a second-order high-pass filter is:

[0045] make The frequency response expression is obtained:

[0046]

[0047] In the formula, A γ is the frequency response, which describes the relationship between the input signal frequency and the output signal frequency; f 0 is the cut-off frequency; Q is the quality factor, which indicates the selectivity and bandwidth of the filter; f is the input signal frequency; j is the imaginary unit, which indicates the phase difference.

[0048] Furthermore, the sound processing module also includes a high-speed analog-to-digital converter for cooperating with a second-order active low-pass filter and a second-order high-pass filter.

[0049] The beneficial effects of the present invention are:

[0050] 1. Multi-dimensional monitoring:

[0051] The abnormal sound recognition and detection device for high-voltage equipment in the substation integrates multiple technologies such as sound recognition, ozone recognition and partial discharge test data recognition, realizes multi-dimensional monitoring of the status of high-voltage equipment, and improves the accuracy and reliability of fault identification.

[0052] 2. Real-time:

[0053] The abnormal sound recognition and detection device for high-voltage equipment in the substation can detect the sound print, ozone content, and partial discharge data of the high-voltage equipment in operation and issue an alarm in a timely manner according to on-site needs, promptly discover potential safety hazards, and provide timely warning information for operation and maintenance personnel.

[0054] 3. Visualization:

[0055] The abnormal sound recognition and detection device for high-voltage equipment in the substation uses acoustic imaging technology to convert sound signals into visual images. At the same time, it converts parameters such as amplitude, frequency and phase of partial discharge detection data into visual images, allowing operation and maintenance personnel to intuitively understand the sound distribution of the equipment and facilitate the location and troubleshooting of fault points.

[0056] 4.Intelligent:

[0057] The abnormal sound recognition and detection device for high-voltage equipment in the substation adopts a machine learning algorithm, which can automatically identify abnormal sound patterns and partial discharge signals, reducing the workload of operation and maintenance personnel and improving work efficiency.

[0058] 5. Application prospects:

[0059] The abnormal sound recognition and detection device for high-voltage equipment in substations provides a new technical means for the operational safety and stability of the power industry. The device can be widely used in the monitoring of high-voltage equipment in various substations to achieve real-time monitoring and early warning of abnormal sounds of equipment. By timely discovering and handling equipment failures, the risk of equipment damage and power outages can be effectively reduced, and the reliability and economy of the power system can be improved.

[0060] In addition, with the continuous development of artificial intelligence and big data technology, the abnormal sound recognition and detection device for high-voltage equipment in substations is expected to achieve more intelligent monitoring and analysis. For example, by learning and analyzing historical data, the sound recognition algorithm and judgment logic are continuously optimized to improve the performance and accuracy of the device; at the same time, the device is combined with intelligent inspection robots, drones and other technologies to achieve comprehensive automated monitoring and maintenance of high-voltage equipment in substations. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a schematic diagram of module connections of the abnormal sound recognition and detection device for high-voltage equipment in a substation according to the present invention;

[0062] Figure 2 This is a schematic diagram of the connection of various modules of the voiceprint sensing unit of the present invention;

[0063] Figure 3 A model diagram of a second-order low-pass filter of the sound processing module of the present invention;

[0064] Figure 4 and Figure 5 is the amplitude-frequency characteristic curve of the second-order low-pass filter;

[0065] Figure 6 This is the model diagram of the second-order high-pass filter;

[0066] Figure 7 is the waveform after Fourier transformation;

[0067] Figure 8 is a process diagram after wavelet transformation;

[0068] Fig. 9 The final graph is formed. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings of 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without creative work are within the scope of protection of the present invention.

[0070] The traditional manual inspection method is not only inefficient, but also endangers the personal safety of operation and maintenance personnel and is difficult to detect potential equipment hazards; therefore, this embodiment provides a substation high-voltage equipment abnormal sound recognition and detection device, such as Figure 1 As shown, the abnormal sound recognition and detection device for high-voltage equipment in a substation includes a voiceprint recognition unit, an ozone gas detection unit, a partial discharge sensor unit, a data processing and analysis unit, and an early warning unit;

[0071] Among them, the voiceprint recognition unit, the ozone gas detection unit, and the partial discharge sensor unit are all connected to the data processing and analysis unit for data transmission; the voiceprint recognition unit collects the operating sound of the equipment in real time; the ozone gas detection unit detects the ozone concentration in the air; the partial discharge sensor unit collects the partial discharge signal of the equipment; the data processing and analysis unit processes and analyzes the operating sound data, ozone concentration data, and partial discharge signal of the equipment, and the data processing and analysis unit is controlled and connected with the early warning unit to realize accurate identification and early warning of abnormal status of the equipment based on the analysis results.

[0072] Voiceprint recognition unit:

[0073] The voiceprint recognition unit is the core part of this device. It uses acoustic imaging to convert sound signals into visual images. By collecting, analyzing and processing the sound signals generated during the operation of high-voltage equipment, the operation and maintenance personnel can intuitively understand the sound distribution of the equipment, thereby judging the operating status of the equipment and whether there are any abnormalities. At the same time, combined with machine learning algorithms, it can automatically identify abnormal sound patterns and accurately locate the fault point.

[0074] Feature extraction of sound signals is a key step in abnormal sound recognition. By performing spectrum analysis, time domain analysis, and waveform analysis on sound signals, characteristic parameters that reflect the operating status of the device can be extracted. For example, by analyzing the frequency, amplitude, waveform, and other characteristics of the sound, it can be determined whether the device is abnormal. Figure 2 As shown, the voiceprint sensing unit includes four parts: a sound collection module, a sound processing module, a feature extraction module and a pattern recognition module. The sound collection module is connected to the sound processing module for data transmission, the sound processing module is connected to the feature extraction module for data transmission, and the feature extraction module is connected to the pattern recognition module for data transmission.

[0075] The sound collection module is installed on the high-voltage equipment of the substation, and uses a high-sensitivity microphone array as a voiceprint sensor. The voiceprint sensor has a sensitivity of 50mV / Pa, a frequency range of 6.3 to 20kHz, and a noise floor of <33dB. It is responsible for collecting abnormal sounds from the high-voltage equipment of the substation. Using a high-sensitivity microphone array as a voiceprint sensor can capture the weak sounds generated during the operation of the equipment.

[0076] The sound processing module is installed in the data processing and analysis unit. The sound processing module pre-processes the collected sound signals, including denoising, filtering and enhancement. Denoising can effectively remove environmental noise and background noise and improve the signal-to-noise ratio of the sound signal; filtering can filter out high-frequency and low-frequency noise and retain useful sound features; enhancement can further improve the clarity and recognition of the sound signal.

[0077] In order to eliminate the influence of interference signals, a filtering circuit is added to the circuit of the sound processing module. The filtering circuit consists of a second-order active low-pass filter and a second-order high-pass filter, as follows:

[0078] Second-order low-pass filter:

[0079] The characteristic of a second-order low-pass filter is that signals below the cutoff frequency pass through, and signals above the cutoff frequency are attenuated 100 times for every 10 times higher frequency. Figure 3 It is a model of a second-order low-pass filter.

[0080] for Figure 3 The transfer function of the second-order low-pass filter shown is:

[0081]

[0082] Where G(s) is the filter transfer function; U 0 (s) is the Laplace transform of the output signal; U i (s) is the Laplace transform of the input signal; G 0 Center frequency gain, which indicates the gain of the filter at the center frequency; ω m is the natural angular frequency, which represents the natural frequency of the system; ξ is the damping coefficient, which determines the response characteristics of the system; S is the complex frequency variable, which represents the frequency domain characteristics of the signal.

[0083] Center frequency gain In the formula, R r and R are Figure 3 The resistance value of the resistor.

[0084] The natural angular frequency is In the formula, R 1 , R 2 , C 1 , C 2 They are Figure 3 The resistance value of the resistor and the capacitance value of the capacitor;

[0085] Damping coefficient

[0086] Using the frequency normalization method, the transfer function of the above second-order low-pass filter is:

[0087]

[0088] Where λ is the normalized frequency parameter, which is usually related to the characteristic frequency of the system; G m is the normalized gain, usually set to 1 during the normalization process.

[0089] like Figure 4 and Figure 5 As shown, Figure 4 and Figure 5 The amplitude-frequency characteristic curve of the second-order low-pass filter is shown in Figure 2. The slope of the stopband attenuation characteristic is -40dB / 10oct. The various parameters of the second-order low-pass filter affect its filtering characteristics. For example, the size of the damping coefficient determines whether the amplitude-frequency characteristic has a peak or the height of the resonance peak.

[0090] The amplitude-frequency characteristic curve of the second-order low-pass filter:

[0091] The quality factor of the second-order low-pass filter circuit diagram shows the state of the filter passband, from which we can get:

[0092]

[0093] When the frequency of the input signal reaches the cutoff frequency f = 1 / 2πRC, the filter begins to attenuate and the amplitude of the output signal is approximately half of the input signal; therefore, when appropriate values ​​are selected for resistors and capacitors, the frequency response of the signal can be effectively controlled.

[0094] Second-order high-pass filter:

[0095] The characteristic of a second-order high-pass filter is that signals above the cutoff frequency pass through, and signals below the cutoff frequency are attenuated 100 times for every 10 times higher frequency; Figure 6 As shown, Figure 6 It is an ideal physical model of a second-order high-pass filter; as can be seen from the figure, the model of a second-order high-pass filter consists of two RC filter circuits and a common-mode proportional amplifier circuit, and its characteristics are high input impedance and low output impedance.

[0096] The second-order high-pass filter has a bandpass gain of:

[0097]

[0098] In the formula, A γp is the passband gain; R f is the feedback resistor; R 1 is the input resistance;

[0099] The transfer function of a second-order high-pass filter is:

[0100]

[0101] In the formula, A γ (s) represents the frequency response transfer function; s is the complex frequency variable; C is the capacitance value; R is the resistance value;

[0102] The frequency response of a second-order high-pass filter is:

[0103] make The frequency response expression is obtained:

[0104]

[0105] In the formula, A γ is the frequency response, which describes the relationship between the input signal frequency and the output signal frequency; f 0 is the cut-off frequency; Q is the quality factor, which indicates the selectivity and bandwidth of the filter; f is the input signal frequency; j is the imaginary unit, which indicates the phase difference.

[0106] According to the above calculation formula and design requirements, Q=0.7, the upper cutoff frequency is 1.1KHz, the 22nd harmonic, and the lower cutoff frequency is 35Hz, 0.7 times the fundamental frequency. The high-pass filter resistor is 5K, the capacitor is 10nF, and the feedback amplifier resistor is 10K and 12K respectively; the low-pass filter resistor is 330K, the capacitor is 100nF, and the feedback amplifier resistor is 10K and 1.5K respectively.

[0107] Furthermore, in order to cooperate with the second-order active low-pass filter and the second-order high-pass filter, the sound processing module also includes a high-speed analog-to-digital converter. In the analog-to-digital converter ADC, the characteristics of the input current are affected by many factors, including sampling rate, input voltage and signal source impedance. The analog input terminal has a sampling period (called t ACQ ) is charged to a built-in capacitor. The following is a comprehensive analysis of the analog input and conversion process:

[0108] Input current and sampling period:

[0109] The input current is only in the sampling period t ACQ The source impedance must be able to charge the different input capacitances (10pF or 20pF) at the maximum sampling rate (280ns).

[0110] Input impedance and conversion period:

[0111] During the conversion cycle, the input impedance is greater than 1MΩ and there is no input current. To ensure the start-up condition, the sampling capacitor needs to be charged to the fixed internal reference voltage before entering the sampling mode.

[0112] Input voltage range:

[0113] The input voltage must be kept within the specified range: [HVSS–0.2V, HVDD+0.2V] to ensure linear conversion.

[0114] Driving operational amplifier bandwidth requirements:

[0115] The minimum –3dB bandwidth is given by:

[0116]

[0117] In the formula, f -3dB is the minimum -3dB bandwidth of the operational amplifier; n is the resolution of the system, usually specified as the number of bits of the digital signal. If the ADS8556 high-speed analog-to-digital converter is used, the resolution is 16 bits, that is, n = 16; t ACQ Sampling time; In(2) is a constant of the natural logarithm and has a value of approximately 0.693.

[0118] For the minimum sampling time taCQ =280ns, the minimum bandwidth required is 6.7MHz; if longer sampling times are allowed, the bandwidth requirement can be reduced.

[0119] Source impedance requirements:

[0120] The driving op amp can be omitted if the signal source impedance satisfies the following conditions:

[0121]

[0122] In the formula, R SOUrCE is the impedance of the signal source, which indicates the resistance of the signal source to the system; Cs is the input sampling capacitor, which is usually used to form an RC time constant with the signal source; R SER It is a series resistor, usually connected between the signal source and the input node; R SW It is the switch resistance, usually the resistance associated with the signal path during the signal sampling process.

[0123] Where, Cs = 10pF, R SER =200Ω, R SW =130Ω; for t ACQ =280ns, source impedance should be less than 2.0kΩ, and the source impedance range can be expanded for long sampling time.

[0124] Conversion process control:

[0125] The analog input (CH_x0 / 1) of each channel group remains valid on the rising edge of the corresponding CONVST_x signal. In software mode (except sequential mode), all 6 ADCs share the CONVST_A signal; the conversion will start on the next edge of the conversion clock. During the conversion, CONVST_x should continue to be high to keep the BUSY signal valid; the falling edge of the conversion process puts the related ADC into power-down mode.

[0126] BUSY signal and data reading:

[0127] The BUSY signal indicates the conversion status, which goes high on the rising edge and goes low when the output data is available. When the BUSY signal goes low, the data can be read immediately. During conversion, the falling edge of CONVST_x causes the ADC to enter power-down mode; while in continuous mode, the BUSY signal decreases by one clock cycle.

[0128] In summary, the performance and stability of ADC are affected by multiple factors such as input voltage, source impedance, sampling time and signal control. By properly configuring these parameters, a fast and accurate signal conversion process can be ensured.

[0129] The feature extraction module is installed in the data processing and analysis unit. The feature extraction module extracts characteristic parameters reflecting the equipment status from the processed sound signal. The interference signal frequency is basically between 0 and 20kHz, while the vibration and background noise signal spectra are mainly concentrated below 10kHz. By observing the distribution of each noise spectrum, the noise category can be further distinguished, such as the noise generated by magnetostriction (mainly 100Hz base frequency, multiplied), the noise caused by DC bias magnetization (a series of high-order and odd-order harmonics), abnormal looseness, fan noise, etc. The detection of sound waves above 20kHZ can effectively detect the surface discharge, suspension discharge model, corona discharge, etc. of the equipment. These characteristic parameters include the spectral characteristics, time domain characteristics and statistical characteristics of the sound. Through appropriate feature extraction methods, the sound signal can be effectively converted into numerical data that can be processed by the computer, providing a basis for subsequent pattern recognition.

[0130] The pattern recognition module is installed in the data processing and analysis unit; after the collected data is processed and calculated, the final result of the function is presented in the form of a "spectrum-energy" spectrum. Judgment method: The pattern recognition module uses a machine learning algorithm to learn and classify the extracted characteristic parameters. The pattern recognition module uses deep learning to build a model that can automatically learn the mapping relationship between sound characteristics and device status; by learning a large number of normal and abnormal sound samples, the model can gradually master the ability to identify abnormal sounds of the device. In actual applications, when the sound acquisition module collects a new sound signal, the pattern recognition module will analyze and process it and determine whether the device is abnormal. When the device is abnormal, its sound characteristics will change. By comparing the voiceprint in the normal state with the voiceprint in the abnormal state, the abnormal situation of the device can be discovered in time. After extracting the characteristic parameters of the sound signal, the device uses an abnormal sound recognition algorithm to classify and identify the characteristic parameters. Commonly used abnormal sound recognition algorithms include algorithms based on pattern recognition, algorithms based on machine learning, and algorithms based on deep learning. Through learning and training a large number of samples, accurate recognition of abnormal sounds is achieved.

[0131] Furthermore, this embodiment provides a specific method in which the voiceprint recognition unit performs spectrum analysis and time domain analysis on the sound signal:

[0132] Time domain analysis: that is, the waveform of the sound. By observing whether there are regular signals in the waveform, it is generally used in combination with subsequent spectrum and time-spectrum spectrum analysis.

[0133] Spectral analysis: obtained by fast Fourier transform of sound waveform. Fourier transform formula:

[0134] F( ω )=∫f(t)e-iwtdt

[0135] Where F(ω) is the Fourier transform result; f(t) is the original signal or time domain function; e -iwt It is a complex function related to frequency w and time t. It is used as a basis function in Fourier transform to represent the basic components of the frequency domain signal. i is the imaginary unit; dt represents the differential of the integral variable, and t is the integral variable.

[0136] This formula represents the magnitude relationship of each frequency component in the signal; in the analysis of the transformer's operating status, its health status can be evaluated by calculating the proportion of the double frequency components of 50Hz and 100Hz.

[0137] Time-frequency spectrum analysis: The sound waveform is transformed by wavelet transform. Wavelet transform formula:

[0138]

[0139] In the formula, WT(a,τ) represents the result of wavelet transform; f(t) is the original signal, which refers to the signal in the time domain; ψ is the wavelet function, which is used to analyze the local characteristics of the signal; a is the scale parameter, which is used to control the expansion degree of the wavelet; τ is the translation parameter, which is used to control the movement of the wavelet function on the time axis; t is the integral variable; is the regularization factor, which is used to adjust the amplitude of wavelet transform.

[0140] This formula expresses the relationship between the frequency components of a sound waveform and time. It can be regarded as a fast Fourier transform of the sound waveform in each infinitely subdivided time period. It is mainly used to distinguish non-periodic signals.

[0141] Vibration fingerprint refers to the vibration characteristics of equipment (such as transformers) under specific working conditions. This feature can be used to identify the health status and operating efficiency of the equipment. The vibration fingerprint can be obtained by extracting the wavelet ridge from the time-frequency spectrum and then converting it into polar coordinates. It is mainly used for comparison with historical data. When the same transformer is subjected to vibration measurement at the same measuring point under the same load, temperature and other industrial control conditions, the obtained vibration fingerprint should maintain a similar form; this means that under normal conditions and the same operating conditions, the vibration characteristics of the transformer should be stable and will not change significantly; if the vibration fingerprint changes significantly, this may indicate that there is a problem with the equipment, such as wear, failure or other abnormal conditions. Therefore, such analysis can be used for equipment health monitoring and fault diagnosis.

[0142] Discrete Fourier Transform (DFT):

[0143] Discrete Fourier Transform (DFT) is often used in digital signal processing to obtain the frequency domain characteristics of the signal.

[0144] The calculation formula of DFT is:

[0145]

[0146] Where X(k) represents the complex value corresponding to the frequency index k of the discrete frequency domain signal after DFT; x(n) represents the sample value corresponding to the time index n of the discrete signal in the time domain; N is the length of the signal, that is, the total number of samples in the signal x(n); n is the sample index of the time domain signal; k refers to the index in the frequency domain; W N is the root of unity of a complex number, used as a basis function in DFT calculations; j represents the imaginary unit.

[0147] The computational complexity of DFT is:

[0148] According to the given DFT calculation formula, we can know that each calculation of a frequency point X(k) requires N complex multiplications and N-1 complex additions, and calculating X(k) of N points requires N2 complex multiplications and N*(N-1) complex additions. When x(n) is a real number, calculating the DFT of N points requires 2*N2 real multiplications and 2*N*(N-1) real additions.

[0149] W N Nature:

[0150] symmetry:

[0151] Periodicity:

[0152] Availability: Where m is the total number of sampling points;

[0153] According to the above properties, the following useful results can be obtained:

[0154]

[0155] Assume that the number of sampling sequence points is N = 2 L , L is an integer. If this condition is not met, we can artificially add several 0s to make the number of sampling sequence points meet this requirement. First, we divide the sequence x(n) into two groups according to odd and even numbers as follows:

[0156]

[0157] Where N represents the total number of points in the sampling sequence; L is an integer; r is the index variable; x(2r) represents the even-indexed sample in the sequence; x(2r+1) is defined as the sequence of even-indexed samples; x 2 (r) is defined as the sequence of odd samples; x 1 (r) is defined as a sequence of even-numbered samples.

[0158] So according to the DFT calculation formula:

[0159]

[0160] At this point, the DFT of an N-point is transformed into a form. At this time, the value of k is 0 to N-1. Now we will discuss it in two parts. When k is 0 to N / 2-1, because x 1 (r)), x 2 (r) is a sequence of N / 2 points, so the above formula can be written as:

[0161]

[0162] When k is N / 2 to N-1, k is replaced by k'+N / 2, and k' is 0 to N / 2-1.

[0163] Based on the above derivation, we can draw the following conclusion: an N-point DFT transformation process can be represented by two N / 2-point DFT transformation processes. The specific formula is the iterative formula of the DFT fast algorithm as shown below:

[0164]

[0165] Where, X N (k) represents the result of N-point DFT; X' N / 2 (k') is the discrete Fourier transform of the even-numbered branch; X" N / 2 (k”) is the discrete Fourier transform of the odd-numbered branches; is the rotation factor.

[0166] By performing even-odd decomposition on DFT, the computational complexity can be effectively reduced, and the calculation of DFT can be converted into a series of smaller problems, thereby realizing the fast Fourier transform FFT, thereby greatly improving the efficiency of signal processing.

[0167] Apply Fourier and wavelet transform to decompose and reconstruct the signal data so that it can be processed and restored into a better signal than the original one. The basis of Fourier transform is sine curves of different frequencies, so Fourier transform decomposes the signal wave into the superposition of sine waves of different frequencies, while wavelet transform decomposes a signal into a series of wavelets. The waveform after Fourier transform is shown in the figure below. Figure 7 As shown; and then through the wavelet transform process, as Figure 8 As shown, the final formation is as Fig. 9 The graph shown enables operation and maintenance personnel to intuitively understand the sound distribution of the equipment, thereby judging the operating status of the equipment and whether there are any abnormalities, as well as automatically identifying abnormal sounds and accurately locating the fault point.

[0168] Ozone gas detection unit:

[0169] During the operation of high-voltage equipment, there is a potential risk of partial discharge. The equipment has corona discharge, tip discharge, arc, etc., which ionizes the oxygen molecules in the air to form ozone molecules, and produces a large number of high-energy electrons and free radicals. These particles react with the oxygen in the air to produce the decomposition product "ozone". The increase in ozone concentration may not only cause oxidation reactions on the surface of the equipment and accelerate the aging of the equipment, but also pose a threat to the health of the staff. People who are exposed to high-concentration ozone for a long time may experience symptoms such as headaches, eye pain, and difficulty breathing.

[0170] The ozone gas detection unit uses an ozone sensor to monitor the ozone concentration around the equipment in real time; the ozone sensor has a resolution of 0.1ppm, a measurement range of 0-1000ppm, and a response time of ≤20 seconds. It communicates with the data processing and analysis unit wirelessly.

[0171] Furthermore, the data processing and analysis unit determines whether there is a discharge fault in the magnetic field space of the equipment. When the ozone concentration exceeds the preset safety threshold, the data processing and analysis unit controls the early warning unit to sound an alarm, reminding the operation and maintenance personnel to stay away from the faulty equipment and area to ensure personal safety.

[0172] Partial discharge sensor unit:

[0173] Partial discharge is one of the common fault types of high-voltage equipment. When high-voltage equipment discharges, due to excessive electric field strength or reduced insulation performance, the charge will break through the insulation layer and discharge. The charge movement, collision and ionization generated during the discharge process will emit a specific sound, accompanied by the generation of partial discharge. Affected by the internal and external environment, partial discharge will cause insulation breakdown if it is not controlled, and may even cause a short circuit explosion accident of high-voltage equipment. In the partial discharge test, the discharge sound is usually sharp and harsh noise, and the sound intensity is relatively high. At the same time, the partial discharge test data will show the amplitude, frequency and phase of the discharge pulse and other parameters. However, the phenomenon of loosening of equipment bolts caused by electrodynamics will not cause discharge itself. Therefore, in the partial discharge test, there is a clear difference between the sound of loosening of bolts and the sound of discharge. The sound of loosening of bolts and other phenomena is usually low, continuous knocking or vibration sound, and the sound intensity is relatively small. The phenomenon of loosening of bolts and other phenomena will not cause obvious discharge pulses. By collecting and analyzing the discharge pulses, it can be judged whether the equipment has discharge.

[0174] The partial discharge sensing unit adopts an air-type ultrasonic partial discharge sensor, whose peak sensitivity is not less than 60dB (V / (m / s)), the average sensitivity is not less than 40dB (V / (m / s)), the detection frequency band is 20kHz~60kHz, the linearity error is not more than ±20%, and the measurement range is -7~73dBuV. The partial discharge sensing unit collects important electrical quantities of the equipment operating status such as the partial discharge characteristic spectrum of each interval of high-voltage equipment and the soundprint characteristic spectrum, and transmits them to the data processing and analysis unit. The data processing and analysis unit displays the test results in the form of continuous amplitude spectrum, PRPD spectrum, etc., monitors the partial discharge of the equipment in real time, determines whether the equipment has insulation defects or hidden faults, effectively identifies and processes partial discharge signals, provides effective fault warning and positioning information, and provides a reliable basis for the implementation of condition maintenance.

[0175] Furthermore, the abnormal sound recognition and detection device for high-voltage equipment in a substation provided in this embodiment includes a detector host, a data processing and analysis unit, an early warning unit, and a sound processing module, a feature extraction module and a pattern recognition module of the voiceprint recognition unit are all installed in the detector host; wherein the data processing and analysis unit is a working machine with a microprocessor; and the early warning unit is an audible and visual alarm.

[0176] The sound collection module of the voiceprint recognition unit, the ozone gas detection unit, the partial discharge sensor unit and other corresponding sensors communicate with the detector host wirelessly, with a data transmission speed of about 10KB / S; the detector host has a built-in lithium battery to power the data processing and analysis unit, the early warning unit, the sound processing module, the feature extraction module and the pattern recognition module, so that it can work continuously for more than 6 hours. The detector host is equipped with three interfaces: a USB host mode interface for USB data export, program upgrade, and device charging; a 2.4G wireless communication interface using a 2dB antenna for environmental sensor module signal access and phase synchronization signal access; and a backup interface that can be used for external display screens for image display, etc.

[0177] In summary, this embodiment provides a substation high-voltage equipment abnormal sound recognition and detection device based on sound recognition, ozone recognition and partial discharge test data recognition. The device has the advantages of multi-dimensional monitoring, real-time, visualization and intelligence, and can achieve effective monitoring of the status of high-voltage equipment and fault warning. Actual tests have shown that the device can significantly improve the operating efficiency and safety of the power grid and reduce operation and maintenance costs.

[0178] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A device for identifying and detecting abnormal sound of high-voltage equipment in a substation, characterized in that: It includes a voiceprint recognition unit, an ozone gas detection unit, a partial discharge sensor unit, a data processing and analysis unit and an early warning unit; the voiceprint recognition unit, the ozone gas detection unit and the partial discharge sensor unit are all connected to the data processing and analysis unit for data transmission; the voiceprint recognition unit collects the operating sound of the equipment in real time; the ozone gas detection unit detects the ozone concentration in the air; the partial discharge sensor unit collects the partial discharge signal of the equipment; the data processing and analysis unit processes and analyzes the operating sound data, ozone concentration data and partial discharge signal of the equipment, and the data processing and analysis unit is controlled and connected to the early warning unit, and accurately identifies and warns the abnormal state of the equipment based on the analysis results.

2. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 1 is characterized in that: The voiceprint sensing unit includes four parts: a sound collection module, a sound processing module, a feature extraction module and a pattern recognition module. The sound collection module is connected to the sound processing module for data transmission, the sound processing module is connected to the feature extraction module for data transmission, and the feature extraction module is connected to the pattern recognition module for data transmission.

3. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 1 is characterized in that: The ozone gas detection unit adopts an ozone sensor.

4. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 1, characterized in that: The partial discharge sensing unit adopts an air-type ultrasonic partial discharge sensor.

5. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 1, characterized in that: The abnormal sound recognition and detection device for high-voltage equipment in a substation includes a detector host, a data processing and analysis unit, an early warning unit, and a sound processing module, a feature extraction module and a pattern recognition module of a voiceprint recognition unit are all installed in the detector host; the data processing and analysis unit is a working machine with a microprocessor; the early warning unit is an audible and visual alarm; the sound collection module, the ozone gas detection unit, the partial discharge sensor unit and other corresponding sensors of the voiceprint recognition unit communicate with the detector host wirelessly; the detector host has a built-in lithium battery to power the data processing and analysis unit, the early warning unit, the sound processing module, the feature extraction module and the pattern recognition module; the detector host is equipped with a USB main mode interface, a 2.4G wireless communication interface and a backup interface.

6. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 1, characterized in that: The voiceprint recognition unit performs spectrum analysis and time domain analysis on the sound signal as follows: Time domain analysis: the waveform of the sound, by observing whether there are regular signals in the waveform, combined with subsequent spectrum and time-spectrum analysis; Spectral analysis: It converts the time domain signal into frequency domain representation and analyzes its frequency components. This conversion is achieved through fast Fourier transform FFT. The specific Fourier transform formula is: Where F(ω) is the Fourier transform result; f(t) is the original signal or time domain function; e -iwt It is a complex function related to frequency w and time t. It is used as a basis function in Fourier transform to represent the basic components of the frequency domain signal. i is an imaginary unit. dt represents the differential of the integral variable, and t is the integral variable. Time-frequency spectrum analysis: Use wavelet transform to obtain the time-frequency characteristics of the signal. The formula is: In the formula, WT(a,τ) represents the result of wavelet transform; f(t) is the original signal, which refers to the signal in the time domain; ψ is the wavelet function, which is used to analyze the local characteristics of the signal; a is the scale parameter, which is used to control the expansion degree of the wavelet; τ is the translation parameter, which is used to control the movement of the wavelet function on the time axis; t is the integral variable; is the regularization factor, which is used to adjust the amplitude of wavelet transform; Discrete Fourier Transform DFT: DFT is an important tool for obtaining the frequency domain characteristics of signals. Its calculation formula is: Where X(k) represents the complex value corresponding to the frequency index k of the discrete frequency domain signal after DFT; x(n) represents the sample value corresponding to the time index n of the discrete signal in the time domain; N is the length of the signal, that is, the total number of samples in the signal x(n); n is the sample index of the time domain signal; k refers to the index in the frequency domain; W N is the root of unity of a complex number, used as a basis function in DFT calculations; j represents the imaginary unit; The computational complexity of DFT: According to the given DFT calculation formula, each calculation of a frequency point X(k) requires N complex multiplications and N-1 complex additions; Calculating X(k) of N points requires N2 complex multiplications and N*(N-1) complex additions. When x(n) is a real number, calculating the DFT of N points requires 2*N2 real multiplications and 2*N*(N-1) real additions. Decomposition process of discrete Fourier transform: by decomposing the original signal into odd and even values, and using the correlation property of WN, an N-point DFT is transformed into a combination of two N / 2-point DFTs; The waveform after Fourier transformation is then transformed through wavelet transformation to finally form a spectrum, thereby judging the operating status of the equipment and whether there are any abnormalities.

7. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 2, characterized in that: The sound collection module uses a highly sensitive microphone array as a voiceprint sensor to capture the weak sounds generated during the operation of the device; The sound processing module pre-processes the collected sound signals, including: removing environmental noise and background noise to improve the signal-to-noise ratio of the sound signal; filtering out high-frequency and low-frequency noise to retain useful sound features; and improving the clarity and recognition of the sound signal; The feature extraction module extracts feature parameters reflecting the device status from the processed sound signal to identify different types of noise; by analyzing the spectral characteristics of the noise, different noise sources can be distinguished; The pattern recognition module presents the results from the processed data in the form of a graph, uses a machine learning algorithm to learn and classify the extracted feature parameters, and builds a model to automatically learn the relationship between sound features and device status, thereby identifying abnormal device sounds.

8. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 8, characterized in that: Under abnormal circumstances, the sound characteristics will change; the pattern recognition module detects abnormalities in a timely manner by comparing the voiceprints in normal and abnormal states.

9. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 2, characterized in that: The circuit of the sound processing module includes a filter circuit, and the filter circuit includes a second-order active low-pass filter and a second-order high-pass filter; The transfer function of a second-order low-pass filter is: Where G(s) is the filter transfer function; U0(s) is the Laplace transform of the output signal; U i (s) is the Laplace transform of the input signal; G0 is the center frequency gain, which represents the gain of the filter at the center frequency; ω m is the natural angular frequency, which indicates the natural frequency of the system; ξ is the damping coefficient, which determines the response characteristics of the system; S is the complex frequency variable, which indicates the frequency domain characteristics of the signal; Using the frequency normalization method, the transfer function of the above second-order low-pass filter is: Where λ is the normalized frequency parameter, which is usually related to the characteristic frequency of the system; G m is the normalized gain, which is usually set to 1 during the normalization process; The model of the second-order high-pass filter includes two RC filter circuits and a common-mode proportional amplifier circuit. It has high input impedance and low output impedance. Signals above the cutoff frequency pass through, and signals below the cutoff frequency are attenuated 100 times for every 10 times higher frequency. The second-order high-pass filter has a bandpass gain of: In the formula, A γp is the passband gain; R f is the feedback resistor; R1 is the input resistor; The transfer function of a second-order high-pass filter is: In the formula, A γ (s) represents the frequency response transfer function; s is the complex frequency variable; C is the capacitance value; R is the resistance value; The frequency response of a second-order high-pass filter is: make The frequency response expression is obtained: In the formula, A γ is the frequency response, which describes the relationship between the input signal frequency and the output signal frequency; f0 is the cut-off frequency; Q is the quality factor, which indicates the selectivity and bandwidth of the filter; f is the input signal frequency; j is the imaginary unit, which indicates the phase difference.

10. The abnormal sound recognition and detection device for high-voltage equipment in a substation according to claim 2, characterized in that: The sound processing module also includes a high-speed analog-to-digital converter for cooperating with a second-order active low-pass filter and a second-order high-pass filter.

Images