Partial discharge positioning method based on super-inductance photoelectric detection

Abstract

The invention discloses a partial discharge positioning method based on super-inductance photoelectric detection, belongs to the technical field of partial discharge detection of power equipment, and effectively overcomes the limitation that a traditional detection technology is difficult to accurately position a partial discharge position and cannot monitor the state of the equipment in real time in an electromagnetic interference environment. Comprising the following steps: attaching three TEV sensors and an ultraviolet pulse sensor to different surfaces of a switch cabinet, and constructing a detection network about the TEV sensors and the ultraviolet pulse sensor; establishing a mathematical model relational expression of the local discharge source coordinates with respect to the positions of the sensors; extracting the time of a signal abrupt change point from the collected signal waveform; and solving the model relational expression to obtain the spatial position of the partial discharge source. According to the method, the super-photosensitive electric signals generated during partial discharge are used as key characteristic parameters for evaluation, the transient voltage-to-ground technology and the ultraviolet pulse detection technology are fused, and accurate space positioning of the partial discharge phenomenon in the switch cabinet can be achieved.

Description

Technical Field

[0001] The present invention relates to a method for locating partial discharge of power system equipment, and in particular to a method for locating partial discharge based on super-sensitive photoelectric detection, belonging to the technical field of partial discharge detection of power equipment. Background Art

[0002] Switchgear is a key power distribution and control equipment used in power systems and is widely deployed in various power conversion, distribution and protection scenarios. In the actual operation of the switchgear, the internal components and insulation system of the switchgear are often in high voltage, high current and complex electromagnetic environment. These conditions can easily cause partial discharge, which seriously affects the performance and life of the switchgear. Therefore, it is of great significance to predict the life of the switchgear by monitoring the relevant status of partial discharge of the switchgear.

[0003] At present, the local discharge detection technology for switch cabinets mainly includes pulse current method, ultrasonic detection method and ultra-high frequency method, but there are still the following problems: 1) The pulse current method is not good enough in terms of anti-interference ability, especially when the switch cabinet needs to be accurately located, its ability is somewhat insufficient, and this method is not suitable for online detection systems. 2) The ultrasonic detection method uses ultrasonic sensors to capture the sound wave signals generated by local discharge, which is mainly used to qualitatively determine the presence or absence of local discharge signals. It cannot effectively determine the discharge type and discharge amount. 3) Although the ultra-high frequency detection method has high detection accuracy, its application depends on the arrangement of the antenna array. In actual operation, due to the limited internal space and complex electrical structure of the switch cabinet, the installation position of the antenna may be restricted, which brings certain challenges to the application of this method.

[0004] Therefore, there is an urgent need to design a partial discharge detection method for switch cabinets, which can quickly and accurately locate the position of the switch cabinet and monitor the equipment status in real time. Summary of the invention

[0005] The purpose of the present invention is to solve the limitations of existing traditional partial discharge detection methods, which are difficult to accurately locate the partial discharge position in an electromagnetic interference environment and cannot monitor the equipment status in real time, and to provide a partial discharge positioning method based on super-sensitive photoelectric detection. The super-sensitive photoelectric signal generated during partial discharge is used as a key characteristic parameter for evaluation, and the two technologies of transient ground voltage and ultraviolet pulse detection are integrated. The signal processing algorithm is combined to extract the signal mutation point time, and a mathematical model is used to construct the relationship between the partial discharge position and the sensor layout. It can perform precise positioning calculations and successfully realize the precise spatial positioning of the partial discharge phenomenon inside the switch cabinet.

[0006] To achieve the above object, the present invention adopts the following technical solution: a partial discharge locating method based on super-sensitive photoelectric detection, comprising the following steps:

[0007] S1. Constructing a detection network for TEV sensors and UV pulse sensors:

[0008] Three TEV sensors and one UV pulse sensor are attached to different surfaces of the switch cabinet, with one sensor fixed on each of the four central sides of the switch cabinet to form a three-dimensional four-corner structure. The switch cabinet is connected to the actual circuit to collect and store the transient voltage waveform to ground and the UV pulse waveform during partial discharge of the switch cabinet.

[0009] S2. Establish the mathematical model of the local discharge source coordinates with respect to the sensor:

[0010] Define point S1 as the location of the UV pulse sensor and its coordinates are (x 1 ,y 1 ,z 1 ), S2, S3 and S4 are the locations of three TEV sensors respectively, and point P is the location of the local discharge source with coordinates (x, y, z); and the local discharge source coordinates P(x, y, z) are established with respect to each sensor location X(x i ,y i ,z i ), the fitness value f is expressed as:

[0011]

[0012] In the formula, the coordinate point (x 1 ,y 1 ,z 1 ) marks the position of the reference UV pulse sensor, (x i ,y i ,z i ) is the spatial position of the remaining sensors, v is the propagation speed of ultraviolet light, t i1 is the time difference between the remaining sensors and the reference UV pulse sensor based on the mutation point;

[0013] S3. Extract the time of the signal mutation point from the collected signal waveform:

[0014] According to the signal waveform collected in step S1, the collected signal waveform is first converted to digital through an analog-to-digital converter (ADC), denoised and filtered based on wavelet transform, and amplified by preprocessing to eliminate interference, and then the envelope of the preprocessed signal waveform is extracted to obtain a smooth curve, and finally the time t of the signal mutation point of the envelope is detected by the Pettitt mutation detection method. iExtraction is performed, and finally the time difference t of the mutation point is calculated. i1 ;

[0015] The specific detection process of the Pettitt mutation detection method is as follows:

[0016] S31) Define dynamic window size:

[0017] For the time dynamic window size used for the RMS value of partial discharge in the switch cabinet, an appropriate dynamic window size is defined according to the statistical characteristics of the actual signal waveform;

[0018] S32) Sliding window:

[0019] The sliding window technology is used to scan the entire time series point by point. During the traversal process, the window will move forward one by one or across multiple time points, and new data points will be included to continuously analyze the changing trends in the series;

[0020] S33) Calculate the RMS value:

[0021] For the waveform data in each time window, calculate the root mean square value (RMS value) of the partial discharge of the switch cabinet. In the formula, x i is a waveform data point, x i is the voltage or current value, N is the number of data points in the window;

[0022] S34) Construct a new time series:

[0023] The RMS value of each window is used as a data point of the new time series. As the sliding window is traversed, a series of RMS values ​​are obtained, which constitute the new time series.

[0024] S35) Apply Pettitt mutation test:

[0025] Apply the Pettitt mutation test to the newly constructed time series based on the RMS value to detect the mutation point in the series. Assume that at this time x i is the voltage, then the Pettitt mutation test formula is as follows:

[0026]

[0027] When the statistic U t,N The absolute value of |U t,N |When the maximum value is obtained, it is the most likely location for the mutation point to appear; at this time: K T =max|U t,N |, the significance level calculation formula is: In the formula, K TRepresents the mutation point. When P < α (using statistical learning or machine learning algorithms to optimize the α value, when α = 0.01 or α = 0.05), K T The time corresponding to the value is the time when the mutation occurs t i ;

[0028] S4. Solve the model relationship to obtain the spatial location of the local discharge source:

[0029] The Brown Bear Algorithm (BOA) is used to solve the model relationship established in step S2, so as to obtain the spatial position of the local discharge source P.

[0030] Furthermore, in step S1, the TEV sensor and the ultraviolet pulse sensor are used to receive the transient voltage to ground and the ultraviolet pulse signal during the partial discharge process, respectively.

[0031] Furthermore, in step S3, the specific process of the preprocessing work is as follows:

[0032] 1) First, set the appropriate reference voltage for the analog-to-digital converter and have a 12-bit or 16-bit resolution;

[0033] 2) After the data is converted by the AD chip, it is processed, uploaded and stored by the FPGA chip. The device continuously collects signals and stores them in the data buffer to prepare for subsequent signal processing;

[0034] 3) In the process of signal processing, multi-layer wavelet decomposition technology is used, 2 to 5 layers are selected to extract information of different frequency bands, and discrete wavelet transform (DWT) is used to decompose the signal to obtain low-frequency and high-frequency components. When processing high-frequency coefficients, threshold denoising technology is applied to effectively eliminate noise and retain useful signal features, while the low-frequency coefficients remain unchanged and the main components of the signal are retained;

[0035] 4) Then, the inverse wavelet transform is used to reconstruct the denoised coefficients to obtain the filtered time domain signal. According to the expected signal amplitude, an operational amplifier with an adjustable gain between 10 and 100 times is selected, and a low-pass filter is added after amplification to eliminate high-frequency noise that may be introduced during the amplification process.

[0036] Furthermore, in step S3, the specific process of extracting the envelope of the preprocessed signal waveform is as follows:

[0037] The preprocessed signal is subjected to Hilbert transform using Python's scipy.signal.hilbert toolkit, and the instantaneous amplitude is calculated based on the Hilbert transform result: Display the envelope of the signal by plotting A(t);

[0038] The Hilbert transform formula is: h(t)=Hibert(x(t)), where x(t) is the preprocessed signal and h(t) is the Hilbert transform result.

[0039] Furthermore, in step S31, the specific process of defining the dynamic window size is:

[0040] 1) Use Python to perform statistics on the preprocessed signal, calculate the statistical characteristics of the signal including the standard deviation and mean, and evaluate the amplitude fluctuation of the signal;

[0041] 2) According to the statistical characteristics of the signal in the current window and the previous window, the rate of change of the signal amplitude is calculated, and the rate of change = |current window mean-previous window mean| / previous window mean;

[0042] 3) According to the result of the rate of change, a threshold is set to decide whether to reduce or increase the window. When the signal changes significantly in a short period of time, the window size is reduced to half of the current window size, while ensuring that the window size is not less than the set minimum window size; when the rate of change is not significant, the window size is increased to twice the current window size.

[0043] Furthermore, in step S4, the solution steps of the Brown Bear Algorithm (BOA) are:

[0044] 1) Initialize the population: The population consists of a group of individuals, each of which represents a potential location where partial discharge occurs;

[0045] 2) Use the fitness value f expression to calculate the fitness value of each individual;

[0046] 3) Select the best individual: In this context, when the fitness value f is closer to zero, it means that this point may be closer to the partial discharge point;

[0047] 4) Update the population: Individuals in the population update their positions through information exchange;

[0048] 5) Repeat steps 2-4 until the maximum number of iterations is reached or other stopping conditions are met, and output the optimal solution found by the algorithm;

[0049] 6) Draw a convergence curve to show the performance changes of the algorithm during the iteration process.

[0050] The beneficial effects of the present invention are:

[0051] 1) The method of the present invention integrates the two technologies of transient ground voltage and ultraviolet pulse detection to realize real-time, multi-modal monitoring of partial discharge phenomena in switch cabinets; by efficiently integrating and analyzing the data waveforms collected by different sensors, it not only enriches the signal acquisition dimension, but also realizes the complementary advantages of the two technical means in signal waveforms, significantly improving the comprehensiveness and accuracy of monitoring.

[0052] 2) In the method of the present invention, a comprehensive preprocessing process is designed for the local discharge signal obtained by the super-sensitive photoelectric detection system. It has multiple functions such as filtering, amplification, and photoelectric conversion. It aims to efficiently process different signals, significantly improve the quality of the signal, enhance the signal-to-noise ratio, and provide a more accurate data basis for subsequent signal analysis and local discharge positioning.

[0053] 3) In view of the limitation that traditional algorithms are prone to fall into local optimality during the positioning process, the present invention introduces a Brown Bear algorithm, which greatly improves the positioning accuracy while increasing the calculation speed. It can lock the specific location of partial discharge more quickly and accurately, and provides strong technical support for equipment maintenance and troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the partial discharge locating method of the present invention;

[0055] Figure 2 A diagram showing the deployment structure of sensors in the method of the present invention;

[0056] Figure 3 Schematic diagram of the structure of the super-sensitive photoelectric partial discharge positioning experimental platform established in an embodiment of the present invention;

[0057] Figure 4 This is an iterative result diagram of the partial discharge position obtained by applying the Brown Bear algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments.

[0059] Example: Figure 3 As shown, the present invention also establishes a super-sensitive photoelectric partial discharge positioning experimental platform developed by LabVIEW, which can input the position coordinates and mutation time of each sensor as part of the initial value into the platform developed by LabVIEW, and display the positioning results on the screen. At the same time, different warnings are given to the staff according to the different degrees of partial discharge.

[0060] like Figure 1 As shown in the flowchart, the present invention provides a method for locating partial discharge based on super-sensitive photoelectric detection, comprising the following steps:

[0061] S1. Constructing a detection network for TEV sensors and UV pulse sensors:

[0062] Three TEV sensors and one UV pulse sensor are attached to different surfaces of the switch cabinet, with one sensor fixed on each of the four central sides of the switch cabinet to form a three-dimensional four-corner structure. The TEV sensor and the UV pulse sensor are used to receive the transient ground voltage and UV pulse signal during partial discharge, respectively. The switch cabinet is connected to the actual circuit to collect and store the transient ground voltage waveform and UV pulse waveform during partial discharge of the switch cabinet.

[0063] S2. Establish the mathematical model of the local discharge source coordinates with respect to the sensor:

[0064] like Figure 2 As shown, point S1 is defined as the location of the UV pulse sensor and its coordinates are (x 1 ,y 1 ,z 1 ), S2, S3 and S4 are the locations of three TEV sensors respectively, and point P is the location of the local discharge source with coordinates (x, y, z); and the local discharge source coordinates P(x, y, z) are established with respect to each sensor location X(x i ,y i ,z i ), the fitness value f is expressed as:

[0065]

[0066] In the formula, the coordinate point (x 1 ,y 1 ,z 1 ) marks the position of the reference UV pulse sensor, (x i ,y i ,z i ) is the spatial position of the other sensors, v is the propagation speed of ultraviolet light, which is the same as the propagation speed of TEV, both 3×10^8m / s, t i1 It is the time difference between the remaining sensors and the benchmark UV pulse sensor based on the mutation point.

[0067] S3. Extract the time of the signal mutation point from the collected signal waveform:

[0068] According to the signal waveform collected in step S1, the collected signal waveform is first converted to digital through an analog-to-digital converter (ADC), denoised and filtered based on wavelet transform, and amplified by preprocessing to eliminate interference, and then the envelope of the preprocessed signal waveform is extracted to obtain a smooth curve, and finally the time t of the signal mutation point of the envelope is detected by the Pettitt mutation detection method. i Extraction is performed, and finally the time difference t of the mutation point is calculated. i1 .

[0069] The specific process of preprocessing is as follows:

[0070] 1) To ensure the accuracy and sensitivity of signal processing, first set the appropriate reference voltage for the analog-to-digital converter and have a 12-bit or 16-bit resolution.

[0071] 2) After the data is converted by the AD chip, it is processed, uploaded and stored by the FPGA chip. The cooperation between the two significantly reduces the data volume and ensures that the system runs quickly and efficiently. The device continuously collects signals and stores them in the data buffer to prepare for subsequent signal processing.

[0072] 3) In the process of signal processing, multi-layer wavelet decomposition technology is adopted, 2 to 5 layers are selected to extract information of different frequency bands, and discrete wavelet transform (DWT) is used to decompose the signal to obtain low-frequency and high-frequency components; threshold denoising technology is applied when processing high-frequency coefficients to effectively eliminate noise and retain useful signal features, while the low-frequency coefficients remain unchanged to ensure that the main components of the signal are retained.

[0073] 4) The denoised coefficients are then reconstructed using an inverse wavelet transform to obtain a filtered time domain signal; an operational amplifier with an adjustable gain between 10 and 100 times is selected based on the expected signal amplitude to ensure that the amplified signal is suitable for subsequent processing; in addition, a low-pass filter is added after amplification to eliminate high-frequency noise that may be introduced during the amplification process.

[0074] The communication module adopts RS485 communication and MAX485 chip, which can not only efficiently transmit the data collected by the centralized unit to the PC through the USB data interface, but also realize efficient data communication.

[0075] The specific process of extracting the envelope of the preprocessed signal waveform is as follows:

[0076] The preprocessed signal is subjected to Hilbert transform using Python's scipy.signal.hilbert toolkit, and the instantaneous amplitude is calculated based on the Hilbert transform result: Display the envelope of the signal by plotting A(t);

[0077] The Hilbert transform formula is: h(t)=Hibert(x(t)), where x(t) is the preprocessed signal and h(t) is the Hilbert transform result.

[0078] The specific detection process of the Pettitt mutation detection method is as follows:

[0079] S31) Define dynamic window size:

[0080] For the time dynamic window size used for the RMS value of partial discharge of the switch cabinet, a suitable dynamic window size is defined according to the statistical characteristics of the actual signal waveform.

[0081] The specific process of defining dynamic window size is:

[0082] 1) Use Python to perform statistics on the preprocessed signal, calculate the statistical characteristics of the signal including the standard deviation and mean, and evaluate the amplitude fluctuation of the signal.

[0083] 2) According to the statistical characteristics of the signal in the current window and the previous window, the rate of change of the signal amplitude is calculated, and the rate of change = |current window mean-previous window mean| / previous window mean.

[0084] 3) According to the result of the rate of change, a threshold is set to decide whether to reduce or increase the window. When the signal changes significantly in a short period of time, partial discharge may have occurred. At this time, the window size is reduced to half of the current window size, thereby increasing the sensitivity to these changes and quickly capturing transient events such as partial discharge. At the same time, ensure that the window size is not less than the set minimum window size to avoid information loss due to the window being too small. If the rate of change is not significant, no partial discharge has occurred at this time, and the window size will be increased to twice the current window size to reduce the impact of random noise on the detection results and enhance stability.

[0085] S32) Sliding window:

[0086] The sliding window technology is used to scan the entire time series point by point. During the traversal process, the window will move forward one by one or across multiple time points, and new data points will be included to continuously analyze the changing trends in the sequence.

[0087] S33) Calculate the RMS value:

[0088] For the waveform data in each time window, calculate the root mean square value (RMS value) of the partial discharge of the switch cabinet. In the formula, x i is a waveform data point, x i is the voltage or current value, and N is the number of data points in the window.

[0089] S34) Construct a new time series:

[0090] The RMS value of each of the above windows is taken as a data point of the new time series. As the sliding window is traversed, a series of RMS values ​​are obtained, which constitute the new time series.

[0091] S35) Apply Pettitt mutation test:

[0092] Apply the Pettitt mutation test to the newly constructed time series based on the RMS value to detect the mutation point in the series. Assume that at this time x i is the voltage, then the Pettitt mutation test formula is as follows:

[0093]

[0094]

[0095] When the statistic U t,N The absolute value of |U t,N |When the maximum value is obtained, it is the most likely location for the mutation point to appear; at this time: K T =max|U t,N |, the significance level calculation formula is: In the formula, K T Represents the mutation point. When P < α (using statistical learning or machine learning algorithms to optimize the α value, when α = 0.01 or α = 0.05), K T The time corresponding to the value is the time when the mutation occurs t i .

[0096] S4. Solve the model relationship to obtain the spatial location of the local discharge source:

[0097] The Brown Bear Algorithm (BOA) is used to solve the model relationship established in step S2, so as to obtain the spatial position of the local discharge source P.

[0098] The Brown Bear Algorithm (BOA) is an intelligent optimization algorithm based on bionics. When solving optimization problems, the algorithm can maintain a balance between exploration and exploitation and does not rely on algorithm-specific parameters. The solution steps of the Brown Bear Algorithm are:

[0099] 1) Initialize the population: The population consists of a group of individuals, each of which represents a potential location where partial discharge occurs;

[0100] 2) Use the fitness value f expression to calculate the fitness value of each individual;

[0101] 3) Select the best individual: In this context, when the fitness value f is closer to zero, it means that this point may be closer to the partial discharge point;

[0102] 4) Update the population: Individuals in the population update their positions through information exchange;

[0103] 5) Repeat steps 2-4 until the maximum number of iterations is reached or other stopping conditions are met, and output the optimal solution found by the algorithm;

[0104] 6) Draw a convergence curve to show the performance changes of the algorithm during the iteration process.

[0105] like Figure 4 As shown in FIG. 1 , the iterative result diagram of the partial discharge position obtained by using the Brown Bear algorithm in an embodiment of the present invention shows the process of finding the optimal solution through multiple iterations. The figure clearly shows that as the number of iterations increases, the algorithm can effectively converge to the optimal partial discharge point position, and specifically quantifies the number of iterations required and the precise value of the final partial discharge position.

[0106] In the local discharge locating method of the present invention, the super-sensitive photoelectric signal generated during local discharge is used as the key characteristic parameter for evaluation. The two technologies of transient voltage to ground and ultraviolet pulse detection are integrated, and the signal mutation point time is extracted by combining the signal processing algorithm. The relationship between the local discharge position and the sensor layout is constructed using a mathematical model. It can perform precise positioning calculations and successfully realize the precise spatial positioning of the local discharge phenomenon inside the switch cabinet.

[0107] The above description is only used to illustrate the technical solution of the present invention rather than to limit it. Other modifications or equivalent substitutions made to the technical solution of the present invention by ordinary technicians in this field should be included in the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for localizing partial discharge based on super-sensitive photoelectric detection, characterized in that: The following steps are involved: S1. Constructing a detection network for TEV sensors and UV pulse sensors: Three TEV sensors and one UV pulse sensor are attached to different surfaces of the switch cabinet, with one sensor fixed on each of the four central sides of the switch cabinet to form a three-dimensional four-corner structure. The switch cabinet is connected to the actual circuit to collect and store the transient voltage waveform to ground and the UV pulse waveform during partial discharge of the switch cabinet. S2. Establish the mathematical model of the local discharge source coordinates with respect to the sensor: Define point S1 as the location of the UV pulse sensor with coordinates (x1, y1, z1), points S2, S3 and S4 as the locations of three TEV sensors, and point P as the location of the local discharge source with coordinates (x, y, z). Use the TDOA algorithm to establish the local discharge source coordinates P (x, y, z) with respect to each sensor location X (x i ,y i ,z i ), the fitness value f is expressed as: In the formula, the coordinate point (x1, y1, z1) is used to mark the position of the reference UV pulse sensor. i ,y i ,z i ) is the spatial position of the remaining sensors, v is the propagation speed of ultraviolet light, t i1 is the time difference between the remaining sensors and the reference UV pulse sensor based on the mutation point; S3. Extract the time of the signal mutation point from the collected signal waveform: According to the signal waveform collected in step S1, the collected signal waveform is first converted to digital through an analog-to-digital converter (ADC), denoised and filtered based on wavelet transform, and amplified by preprocessing to eliminate interference, and then the envelope of the preprocessed signal waveform is extracted to obtain a smooth curve, and finally the time t of the signal mutation point of the envelope is detected by the Pettitt mutation detection method. i Extraction is performed, and finally the time difference t of the mutation point is calculated. i1 ; The specific detection process of the Pettitt mutation detection method is as follows: S31) Define dynamic window size: For the time dynamic window size used for the RMS value of partial discharge in the switch cabinet, an appropriate dynamic window size is defined according to the statistical characteristics of the actual signal waveform; S32) Sliding window: The sliding window technology is used to scan the entire time series point by point. During the traversal process, the window will move forward one by one or across multiple time points, and new data points will be included to continuously analyze the changing trends in the series; S33) Calculate the RMS value: For the waveform data in each time window, calculate the root mean square value (RMS value) of the partial discharge of the switch cabinet. In the formula, x i is a waveform data point, x i is the voltage or current value, N is the number of data points in the window; S34) Construct a new time series: The RMS value of each window is used as a data point of the new time series. As the sliding window is traversed, a series of RMS values ​​are obtained, which constitute the new time series. S35) Apply Pettitt mutation test: Apply the Pettitt mutation test to the newly constructed time series based on the RMS value to detect the mutation point in the series. Assume that at this time x i is the voltage, then the Pettitt mutation test formula is as follows: When the statistic U t,N The absolute value of |U t,N |When the maximum value is obtained, it is the most likely location for the mutation point to appear; at this time: K T =max|U t,N |, the significance level calculation formula is: In the formula, K T Represents the mutation point. When P < α (using statistical learning or machine learning algorithms to optimize the α value, when α = 0.01 or α = 0.05), K T The time corresponding to the value is the time when the mutation occurs t i ; S4. Solve the model relationship to obtain the spatial location of the local discharge source: The Brown Bear Algorithm (BOA) is used to solve the model relationship established in step S2, so as to obtain the spatial position of the local discharge source P.

2. A method for localizing partial discharge based on super-sensitive photoelectric detection according to claim 1, characterized in that: In step S1, the TEV sensor and the ultraviolet pulse sensor are used to receive the transient ground voltage and the ultraviolet pulse signal during the partial discharge process respectively.

3. A method for localizing partial discharge based on super-sensitive photoelectric detection according to claim 1, characterized in that: In step S3, the specific process of pre-processing is as follows: 1) First, set the appropriate reference voltage for the analog-to-digital converter and have a 12-bit or 16-bit resolution; 2) After the data is converted by the AD chip, it is processed, uploaded and stored by the FPGA chip. The device continuously collects signals and stores them in the data buffer to prepare for subsequent signal processing; 3) In the process of signal processing, multi-layer wavelet decomposition technology is used, 2 to 5 layers are selected to extract information of different frequency bands, and discrete wavelet transform (DWT) is used to decompose the signal to obtain low-frequency and high-frequency components. When processing high-frequency coefficients, threshold denoising technology is applied to effectively eliminate noise and retain useful signal features, while the low-frequency coefficients remain unchanged and the main components of the signal are retained; 4) Then, the inverse wavelet transform is used to reconstruct the denoised coefficients to obtain the filtered time domain signal. According to the expected signal amplitude, an operational amplifier with an adjustable gain between 10 and 100 times is selected, and a low-pass filter is added after amplification to eliminate high-frequency noise that may be introduced during the amplification process.

4. The method for localizing partial discharge based on super-sensitive photoelectric detection according to claim 1, characterized in that: In step S3, the specific process of extracting the envelope of the preprocessed signal waveform is as follows: The preprocessed signal is subjected to Hilbert transform using Python's scipy.signal.hilbert toolkit, and the instantaneous amplitude is calculated based on the Hilbert transform result: Display the envelope of the signal by plotting A(t); The Hilbert transform formula is: h(t)=Hibert(x(t)), where x(t) is the preprocessed signal and h(t) is the Hilbert transform result.

5. The method for localizing partial discharge based on super-sensitive photoelectric detection according to claim 1, characterized in that: In step S31, the specific process of defining the dynamic window size is as follows: 1) Use Python to perform statistics on the preprocessed signal, calculate the statistical characteristics of the signal including the standard deviation and mean, and evaluate the amplitude fluctuation of the signal; 2) According to the statistical characteristics of the signal in the current window and the previous window, the rate of change of the signal amplitude is calculated, and the rate of change = |current window mean-previous window mean| / previous window mean; 3) According to the result of the rate of change, a threshold is set to decide whether to reduce or increase the window. When the signal changes significantly in a short period of time, the window size is reduced to half of the current window size, while ensuring that the window size is not less than the set minimum window size; when the rate of change is not significant, the window size is increased to twice the current window size.

6. The method for localizing partial discharge based on super-sensitive photoelectric detection according to claim 1, characterized in that: In step S4, the solution steps of the Brown Bear Algorithm (BOA) are: 1) Initialize the population: The population consists of a group of individuals, each of which represents a potential location where partial discharge occurs; 2) Use the fitness value f expression to calculate the fitness value of each individual; 3) Select the best individual: In this context, when the fitness value f is closer to zero, it means that this point may be closer to the partial discharge point; 4) Update the population: Individuals in the population update their positions through information exchange; 5) Repeat steps 2-4 until the maximum number of iterations is reached or other stopping conditions are met, and output the optimal solution found by the algorithm; 6) Draw a convergence curve to show the performance changes of the algorithm during the iteration process.

Images