Noise reduction method and device for partial discharge signal, detection method and device and transformer system

Abstract

The invention discloses a noise reduction method, a detection method, a device and a system for optimizing a VMDEWT multi-scale partial discharge signal based on CPO. The method comprises the following steps: firstly, obtaining a real partial discharge signal of a power transformer body; performing global optimization on the variation mode decomposition parameter based on CPO to obtain a target decomposition parameter, and performing variation mode decomposition on the real partial discharge signal based on the target decomposition parameter to obtain a plurality of intrinsic mode components; further decomposing the intrinsic mode component through an EWT algorithm and calculating a Pearson's correlation coefficient; and discarding the components lower than the threshold value, and retaining the components higher than the threshold value. Then, the reserved frequency band mode component is reconstructed to obtain a noise reduction signal, and finally, the noise reduction effect is checked through a partial discharge simulation signal. According to the method, the problem of local optimal convergence possibly occurring when an optimal solution is sought by an algorithm can be effectively avoided, it is ensured that population individuals can explore the solution space more comprehensively in the iteration process, and the overall accuracy of partial discharge noise reduction is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of discharge detection, and particularly relates to a noise reduction method, a detection method, a device and a transformer system for multi-scale partial discharge signals optimized by CPO (Crown-Hedgehog Algorithm) for VMD_EWT (VMD is Variational Mode Decomposition; EWT is Empirical Wavelet Transform). Background Art

[0002] Partial Discharge (PD) refers to the discharge phenomenon occurring inside or on the surface of an electrical insulation system. Partial discharge does not immediately cause insulation breakdown, but it indicates that there are defects or deteriorations in the insulation system. Long-term existence will lead to equipment failures and even accidents. Partial discharge mainly occurs in high-voltage equipment. Especially for the main body of a large power transformer, detecting and analyzing partial discharge signals is of great significance for predicting and preventing electrical equipment failures. Through effective detection and analysis of partial discharge, the operation reliability of electrical equipment can be greatly improved, the failure rate can be reduced, the equipment life can be extended, and the safe and stable operation of the power system can be ensured.

[0003] Currently, the detection of partial discharge in the main body of a transformer mainly relies on signals collected by Ultra-High Frequency (UHF), High-Frequency Current Transformer (HFCT) and Acoustic Emission (AE) sensors. However, the main body of a transformer is generally installed outdoors with a harsh environment. Signal acquisition is easily affected by the outside world and is prone to noise interference. Therefore, noise reduction processing is required to reduce noise interference, extract useful discharge information, and conduct subsequent research on partial discharge.

[0004] In the prior art, the noise removal of partial discharge signals mainly adopts wavelet noise reduction methods of soft threshold and hard threshold. Among them, the Hard Thresholding method sets the coefficients with absolute values less than the threshold to zero and retains other coefficients; while the Soft Thresholding method sets the coefficients with absolute values less than the threshold to zero and simultaneously performs threshold shrinkage on the remaining coefficients. Although these wavelet noise reduction methods of hard and soft thresholds can usually provide smoother reconstruction results and reduce the artifact effect, their accuracy and precision are still insufficient, and some useful signal features may be lost.

[0005] Based on the wavelet denoising method with hard and soft thresholds, the "Cable Partial Discharge Signal Denoising Method Based on Variational Mode Decomposition and Concave-Convex Threshold Wavelet" (Modern Electric Power, No. 05, 2022) proposed a denoising method. This method first uses variational mode decomposition (VMD) to decompose and reconstruct the original signal to complete preliminary denoising, and then performs further denoising through wavelet transform with concave-convex thresholds. This method has been tested on simulation signals and measured signals, and the effects have been compared with those of traditional denoising methods. The comparison results prove that this method has a better denoising effect compared with the traditional hard and soft threshold wavelet denoising methods, and the retention effect of partial discharge signal characteristics is also better. However, this method may lead to finding a local optimal solution rather than the global optimal solution, and the current optimal individual position may not necessarily be the global optimal. As the number of iterations increases, the individuals in the population may wrongly gather in the local optimal region, resulting in premature convergence of the algorithm, which will reduce the optimization accuracy of the algorithm and thus affect the overall accuracy of the partial discharge denoising method. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to propose a denoising method, a detection method, a device and a power transformer system for multi-scale partial discharge signals of power transformers based on CPO optimization of VMD_EWT in view of the above deficiencies of the prior art. This method can effectively avoid the problem of local optimal convergence that may occur when the algorithm seeks the optimal solution, ensure that the population individuals can explore the solution space more comprehensively during the iteration process, and thus improve the overall accuracy of the partial discharge denoising method.

[0007] In the first aspect, the present invention provides a denoising method for multi-scale partial discharge signals of power transformers based on CPO optimization of VMD_EWT, and the method includes the following steps:

[0008] Step S1: Obtain the true partial discharge signal of the power transformer body;

[0009] Step S2: Based on the target decomposition parameters, perform variational mode decomposition on the true partial discharge signal to obtain a number of intrinsic mode components;

[0010] Among them, the target decomposition parameters are obtained by globally optimizing the variational mode decomposition parameters with the minimum value of the composite index of residual index and mutual information entropy as the global optimization objective of the fitness function, and using the crown porcupine optimization algorithm;

[0011] Step S3: Further decompose the intrinsic mode components by using the EWT algorithm to obtain frequency band mode components;

[0012] Step S4: Calculate the Pearson correlation coefficient between the frequency band mode components and the intrinsic mode components; and,

[0013] Discard the band modal components with Pearson correlation coefficients lower than the preset threshold, while retaining the band modal components with Pearson correlation coefficients not lower than the preset threshold;

[0014] Step S5: Reconstruct the retained band modal components to obtain the denoised true partial discharge signal, thereby completing the denoising of the partial discharge signal based on the CPO-optimized VMD_EWT multi-scale.

[0015] Furthermore, in the step S2, the global optimization of the variational mode decomposition parameters by using the crowned porcupine optimization algorithm is obtained according to the defense mechanism;

[0016] The defense mechanism specifically includes the first defense mechanism, and / or the second defense mechanism, and / or the third defense mechanism, and / or the fourth defense mechanism;

[0017] The first defense mechanism is visual defense, and its strategy model is as follows:

[0018]

[0019] Where, represents the position of the i-th predator in the (t + 1)-th generation of the first defense mechanism; represents the optimal solution in the t-th generation of the first defense mechanism; represents the position of the i-th predator in the t-th generation of the first defense mechanism; represents the distance between the current porcupine and a randomly selected porcupine in the population; τ 1 , τ 2 represents a normally distributed random number, and the interval is [0, 1];

[0020] The second defense mechanism is sound defense, and its strategy model is as follows:

[0021]

[0022] Where, r 1 , r 2 represent random integers in [1, N]; τ 3 represents a random number between [0, 1]; U 2,1 represents a random probability between [0, 1], represents the position of the i-th predator in the (t + 1)-th generation of the second defense mechanism; represents the position of the i-th predator in the t-th generation of the second defense mechanism; y represents the distance between the current porcupine and a randomly selected porcupine in the population; represents the position of the predator when the random number is r 1 in the t-th generation of the second defense mechanism; represents the position of the predator when the random number is r 2 in the t-th generation of the second defense mechanism;

[0023] The third defense mechanism is that the crested porcupine defends itself by secreting malodorous gases, and its strategic model is as follows:

[0024]

[0025] Among them, r 3 、r 4 、r 5 represent random integers in [1, N]; τ 3 represents a random number between [0, 1]; U 3,1 represents a random probability between [0, 1], represents the position of the i-th predator in the (t + 1)-th generation of the third defense mechanism; represents the position of the i-th predator in the t-th generation of the third defense mechanism; represents the position of the predator with a random number r 3 in the t-th generation of the third defense mechanism; represents the position of the predator with a random number r 4 in the t-th generation of the third defense mechanism; represents the fitness function of the i-th predator in the t-th generation of the third defense mechanism; τ 3 represents a normally distributed random number in the interval [0, 1]; δ 3 represents a random number in the interval [0, 1]; γ 3,t represents a time-related factor;

[0026] The fourth defense mechanism is that the crested porcupine attacks with its own quills, and its strategic model is as follows:

[0027]

[0028] Among them, τ 4 、τ 5 represent random numbers between [0, 1]; represents the position of the i-th predator in the (t + 1)-th generation of the fourth defense mechanism; represents the optimal solution in the t-th generation of the fourth defense mechanism; represents the position of the i-th predator in the t-th generation of the fourth defense mechanism; δ 4 represents a random number in the interval [0, 1]; γ 4,t represents a time-related factor; represents the fitness factor of the i-th predator in the t-th generation of the fourth defense mechanism; α(1 - τ 4 ) + τ 4 represents a dynamic adjustment strategy, where α is a control parameter in the interval [0, 1].

[0029] Furthermore, an improved cyclic population reduction mechanism is established in the defense mechanism;

[0030] The model formula for the improved cyclic population reduction mechanism is as follows:

[0031]

[0032] In the formula, N 1 represents the population size; N min represents the minimum population; N' represents the surrogate population size; t represents the number of iterations; % represents the modulo operation; T max represents the maximum number of iterations; T represents the number of cycles.

[0033] Furthermore, in step S2, the target decomposition parameter is the global optimization objective of the fitness function with the minimum value of the composite index of the residual index and the mutual information entropy. The process of obtaining the fitness function is as follows:

[0034] Step S21: Calculate the residual index index, and the formula is:

[0035]

[0036] where N represents the length of the observed data; y j represents the j-th actual observed value; y' j represents the j-th predicted value;

[0037] Step S22: Calculate the mutual information entropy index, and the formula is:

[0038]

[0039] where p(x,y) represents the joint probability distribution of X and Y; p(x) and p(y) represent the marginal probability distributions of X and Y, X represents the original signal, and Y represents the intrinsic mode component obtained after decomposition;

[0040] Step S23: Weight and combine the residual index index and the mutual information entropy index to obtain a composite index;

[0041] The calculation formula for the composite index CI is;

[0042] CI = λ × RI + β × I;

[0043] λ + β = 1;

[0044] where RI is the residual index index, I is the mutual information entropy index; λ and β are both penalty coefficients, which are positive numbers and λ + β = 1.

[0045] Furthermore, the Pearson correlation coefficient in step S4 has the following calculation formula:

[0046]

[0047] Among them, n represents the number of samples, and n is a natural number greater than 1; pcc j,k represents the correlation coefficient between the j-th component of IMFs and the k-th component of its corresponding EWT decomposition EWT_IMFs; IMFs j,i represents the i-th element of the j-th component of IMFs; EWT_IMFs j,k,i represents IMFs j,i the i-th element of the k-th component of the j-th component.

[0048] Further, the real partial discharge signal is a partial discharge ultra-high frequency signal and / or a pulse current signal;

[0049] In step S2, performing variational mode decomposition on the real partial discharge signal specifically includes:

[0050] Performing variational mode decomposition on the partial discharge ultra-high frequency signal and / or performing variational mode decomposition on the pulse current signal.

[0051] Further, step S1 specifically includes the following steps:

[0052] Step S11: Construct a partial discharge model; and build a partial discharge experimental platform;

[0053] Step S12: Based on the partial discharge experimental platform, adjust the voltage regulator to trigger the partial discharge model to generate a partial discharge signal, and observe the signal pattern through an oscilloscope and a partial discharge detector;

[0054] Step S13: Based on the signal pattern, collect and save the partial discharge pulse signal to complete the acquisition of the real partial discharge signal of the power transformer body.

[0055] Further, after step S5, the method further includes:

[0056] Step S6: Verify the noise reduction effect through a partial discharge simulation signal;

[0057] Step S6 specifically includes the following steps:

[0058] Obtain the power of the partial discharge simulation signal before and after noise reduction respectively;

[0059] Calculate the signal-to-noise ratio SNR between the power of the partial discharge simulation signal before and after noise reduction sim ;

[0060]

[0061] Among them,

[0062] Pow beforeis the power of the partial discharge simulation signal before noise reduction;

[0063] Pow after is the power of the partial discharge simulation signal after noise reduction;

[0064] Evaluate the noise reduction effect according to the signal-to-noise ratio.

[0065] Furthermore, the steps for obtaining the power of the partial discharge simulation signal are as follows:

[0066] Step A1: Obtain the pulse waveform of the partial discharge simulation signal;

[0067] The calculation formula for the pulse waveform of the partial discharge simulation signal is as follows:

[0068]

[0069] where s(t) represents the pulse waveform; A represents the maximum pulse amplitude; Ω1 represents the first pulse attenuation coefficient; Ω2 represents the second pulse attenuation coefficient;

[0070] Step A2: Calculate the reference power P base of the partial discharge simulation signal according to the pulse waveform of the partial discharge simulation signal, and its calculation formula is as follows:

[0071]

[0072] where s i represents the i-th point of the signal; N represents the signal length;

[0073] Step A3: Add Gaussian noise power to the reference power to obtain the power of the partial discharge simulation signal;

[0074] The calculation formula for the power of the partial discharge simulation signal is as follows:

[0075] P total = P base + P noise ;

[0076] where P total represents the power of the partial discharge simulation signal; P noise represents the Gaussian noise power; P base represents the reference power;

[0077] The calculation formula for the Gaussian noise power is as follows:

[0078]

[0079] where randn is standard Gaussian white noise; length is a function that returns the length of the array; SNR input is the preset signal-to-noise ratio of Gaussian noise.

[0080] In a second aspect, the present invention provides a method for detecting partial discharge signals, the method comprising the following steps:

[0081] Adopt the noise reduction method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT described in the first aspect to obtain the true partial discharge signal after noise reduction;

[0082] Detect the true partial discharge signal after noise reduction.

[0083] In a third aspect, the present invention provides a noise reduction device for multi-scale partial discharge signals based on CPO-optimized VMD_EWT, the device comprising:

[0084] An acquisition unit, configured to acquire the true partial discharge signal of the power transformer body;

[0085] A first decomposition unit, connected to the acquisition unit, configured to perform variational mode decomposition on the true partial discharge signal based on target decomposition parameters to obtain a plurality of intrinsic mode components;

[0086] Wherein, the target decomposition parameters are obtained by taking the minimum value of the composite index of the residual index and the mutual information entropy as the global optimization objective of the fitness function, and using the crown porcupine optimization algorithm to globally optimize the variational mode decomposition parameters;

[0087] A second decomposition unit, connected to the first decomposition unit, configured to further decompose the intrinsic mode components by using the EWT algorithm to obtain band mode components;

[0088] A selection unit, connected to the second decomposition unit, configured to discard the band mode components with a Pearson correlation coefficient lower than a preset threshold, and at the same time retain the band mode components with a Pearson correlation coefficient not lower than the preset threshold;

[0089] A reconstruction unit, connected to the selection unit, configured to reconstruct the retained band mode components to obtain the true partial discharge signal after noise reduction, thereby completing the noise reduction of multi-scale partial discharge signals based on CPO-optimized VMD_EWT.

[0090] Further, the device further comprises an inspection unit;

[0091] The inspection unit is configured to inspect the noise reduction effect through partial discharge simulation signals;

[0092] The inspection unit comprises:

[0093] An acquisition module for respectively acquiring the power of the partial discharge simulation signal before and after noise reduction;

[0094] A calculation module, connected to the acquisition module, for calculating the signal-to-noise ratio SNR between the power of the partial discharge simulation signal before and after noise reduction sim ;

[0095] The calculation formula of the signal-to-noise ratio SNR is stored in the calculation module, specifically as follows: sim as follows:

[0096]

[0097] where,

[0098] Pow before is the power of the partial discharge simulation signal before noise reduction;

[0099] Pow after is the power of the partial discharge simulation signal after noise reduction;

[0100] An evaluation module, connected to the calculation module, for evaluating the noise reduction effect according to the signal-to-noise ratio.

[0101] In a fourth aspect, the present invention provides a power transformer system, which includes a memory and a processor. A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the noise reduction method for partial discharge signals based on CPO-optimized VMD_EWT according to the first aspect, or executes the detection method according to the second aspect.

[0102] The present invention realizes multi-scale partial discharge signal noise reduction through CPO optimization of VMD_EWT, which can effectively avoid the problem of local optimal convergence that may occur when the algorithm seeks the optimal solution, ensure that population individuals can explore the solution space more comprehensively during the iteration process, and thus improve the overall accuracy of the partial discharge noise reduction method. The specific beneficial effects are as follows:

[0103] 1. Strong global optimization ability, improving noise reduction accuracy

[0104] The present invention effectively avoids the problem of local optimal convergence that may occur when seeking the optimal solution. Traditional signal processing algorithms are prone to falling into local optimal solutions when optimizing parameters and cannot find the globally optimal noise reduction scheme. By simulating the unique defense and foraging behaviors of the crested porcupine, the crested porcupine optimization algorithm (CPO) enables population individuals to comprehensively explore the solution space from different angles and ranges, increasing the chance of finding the globally optimal solution. This ensures that the VMD parameters can be optimized to the best state, making the decomposition of partial discharge signals more accurate, and thus improving the overall noise reduction accuracy.

[0105] 2. Adaptive adjustment of parameters to enhance the adaptability of signal decomposition

[0106] Through the CPO algorithm, the present invention adaptively adjusts the parameters of VMD according to the characteristics of partial discharge signals. Partial discharge signals are complex and diverse due to different power equipment and operating environments, such as frequency distribution and pulse characteristics. The CPO algorithm can sense the signal characteristics in real time and dynamically adjust the number of modes and penalty factor of VMD. For example, for signals with complex frequency components, CPO can automatically increase the number of modes to decompose the signal more meticulously; for signals with strong noise interference, the penalty factor can be adjusted to improve the anti-interference ability. This adaptive adjustment enables VMD to better adapt to different types of partial discharge signals and effectively separate useful signals and noise components.

[0107] 3. Advantages of multi-scale analysis to comprehensively capture signal characteristics

[0108] By combining VMD and EWT, the present invention realizes multi-scale analysis of signals. VMD decomposes partial discharge signals into a series of intrinsic mode functions (IMFs) with different center frequencies and bandwidths, and these IMFs reflect the main characteristics of the signal at different scales. Empirical wavelet transform (EWT) further processes these IMFs and adaptively constructs a wavelet filter bank based on the frequency characteristics of the signal itself, so as to analyze the local characteristics of the signal more precisely. This multi-scale analysis method can comprehensively capture the characteristic information of partial discharge signals in different frequency ranges and time scales, effectively remove noise interference, and improve the accuracy of positioning and extraction.

[0109] 4. Excellent anti-interference ability to adapt to complex environments

[0110] Through the CPO optimization of the decomposition effect of VMD and the fine processing ability of EWT, the present invention demonstrates excellent anti-interference ability. The VMD optimized by CPO can more accurately separate interference noise from useful signals, and EWT can specifically suppress and remove interference at different scales. For example, EWT effectively filters out interference with frequencies close to partial discharge signals through an adaptive wavelet filter bank, ensuring that the true characteristics of partial discharge signals can still be accurately extracted in a complex electromagnetic environment, thus ensuring the stability and reliability of the noise reduction effect.

[0111] 5. Improve the reliability of fault diagnosis and ensure the safety of the power system

[0112] Accurate denoising of partial discharge signals in the present invention is a key link in the fault diagnosis of power equipment. Partial discharge is an important sign of the degradation of the insulation performance of power equipment and potential faults. Through effective denoising processing, the present invention can provide a more reliable basis for subsequent fault diagnosis. Clear and accurate partial discharge signals can help detect safety hazards in power equipment in a timely manner and take corresponding maintenance and repair measures, thereby ensuring the safe and stable operation of the power system and reducing economic losses and safety risks brought by equipment failures.

[0113] 6. Significantly improve the signal-to-noise ratio

[0114] The present invention can effectively improve the signal-to-noise ratio of partial discharge signals, making important signal features more obvious. This advantage ensures that not only the noise in the denoised signal is effectively suppressed, but also the useful signals are fully retained, further improving the overall processing effect and providing a high-quality data basis for subsequent analysis and diagnosis.

[0115] 7. Adaptive signal processing

[0116] Based on the CPO algorithm, the present invention has the characteristics of adaptive adjustment and can automatically optimize parameters according to different signal characteristics and noise levels. This characteristic enables the method to still maintain excellent performance in the face of diverse and dynamically changing signal environments, further improving the processing effect and accuracy.

[0117] 8. Improve the ability of fault monitoring and prediction

[0118] By significantly improving the quality of partial discharge signals, the present invention provides a more accurate and reliable basis for the health monitoring, fault diagnosis and prediction of electrical equipment. High-quality denoised signals support more accurate analysis and decision-making, helping to improve the operation reliability and safety of the entire power system. Brief description of the drawings

[0119] Figure 1 It is a schematic diagram of the denoising method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT in the embodiment of the present invention;

[0120] Figure 2 It is a schematic flow diagram of the denoising method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT in the embodiment of the present invention;

[0121] Figure 3 It is a detailed flow schematic diagram in the embodiment of the present invention;

[0122] Figure 4 It is a time-domain diagram of the simulated partial discharge pulse signal in the embodiment of the present invention;

[0123] Figure 5The time-domain diagram of the simulated partial discharge pulse signal with a signal-to-noise ratio of 10 in the embodiment of the present invention;

[0124] Figure 6 The time-domain diagram of the simulated partial discharge pulse noisy signal after conventional EMD noise reduction in the embodiment of the present invention;

[0125] Figure 7 The time-domain diagram of the simulated partial discharge pulse noisy signal after conventional EWT noise reduction in the embodiment of the present invention;

[0126] Figure 8 The time-domain diagrams of each mode after CPO_VMD decomposition of the simulated partial discharge pulse noisy signal in the embodiment of the present invention;

[0127] Figure 9 The time-domain diagram of the simulated partial discharge pulse noisy signal after CPO_VMD noise reduction in the embodiment of the present invention;

[0128] Figure 10 The time-domain diagram of the simulated partial discharge pulse noisy signal after CPO_VMD_EWT noise reduction in the embodiment of the present invention;

[0129] Figure 11 The time-domain diagram of the floating discharge pulse signal in the embodiment of the present invention;

[0130] Figure 12 The time-domain diagram of the floating discharge pulse signal after EWT noise reduction in the embodiment of the present invention;

[0131] Figure 13 The time-domain diagrams of each mode of the floating discharge pulse signal after CPO_VMD decomposition in the embodiment of the present invention;

[0132] Figure 14 The time-domain diagram of the floating discharge pulse signal after CPO_VMD_EWT noise reduction in the embodiment of the present invention;

[0133] Figure 15 The fitness curve diagram of the CPO_VMD iteration process in the embodiment of the present invention;

[0134] Figure 16 The schematic diagram of the noise reduction device for multi-scale partial discharge signals based on CPO optimized VMD_EWT in the embodiment of the present invention.

[0135] Reference numerals: 10, acquisition unit; 20, first decomposition unit; 30, second decomposition unit; 40, selection unit; 50, reconstruction unit. Detailed implementation manners

[0136] To enable those skilled in the art to better understand the technical solutions of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0137] It is understood that the specific embodiments and drawings described herein are only for explaining the present invention, rather than limiting the present invention.

[0138] It is understood that, without conflict, the various embodiments in the present invention and the various features in the embodiments may be combined with each other.

[0139] It is understood that for the convenience of description, only the parts related to the present invention are shown in the drawings of the present invention, and the parts unrelated to the present invention are not shown in the drawings.

[0140] It is understood that each unit and module involved in the embodiments of the present invention may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple units and modules may also be integrated into one entity structure.

[0141] It is understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in a different order from that marked in the drawings.

[0142] It is understood that in the flowcharts and block diagrams of the present invention, the possible system architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention are shown. Among them, each block in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified function. Moreover, each block or combination of blocks in the block diagram and flowchart may be implemented by a hardware-based system for implementing the specified function, or may be implemented by a combination of hardware and computer instructions.

[0143] It is understood that the units and modules involved in the embodiments of the present invention may be implemented in software or in hardware. For example, the units and modules may be located in the processor.

[0144] Embodiment 1:

[0145] As Figure 1 shown, this embodiment provides a noise reduction method for optimizing VMD_EWT multi-scale partial discharge signals based on CPO. The present invention aims to accurately process and remove the noise in the partial discharge signals by combining the crown porcupine optimization algorithm, variational mode decomposition, and empirical wavelet transform. This method can adaptively adjust parameters, effectively avoid local optimal convergence, and at the same time realize multi-scale analysis to comprehensively capture signal features, thereby improving the signal-to-noise ratio and enhancing the reliability of fault diagnosis and monitoring. This method is applicable to power systems, electrical equipment health monitoring, and related fields, and can provide strong technical support for the safe operation of equipment.

[0146] Description: The variational mode decomposition (VMD) algorithm is an adaptive signal decomposition method that can decompose complex signals into a set of modes with specific frequencies and bandwidths. It uses the alternating direction method of multipliers for iterative optimization to make each mode have narrowband characteristics. However, the decomposition process is affected by the number of decomposition layers and the penalty factor. If the number of decomposition layers is too large, spurious mode components will be generated; if the number of decomposition layers is small, under-decomposition and mode aliasing will occur, and the noise reduction effect cannot be achieved. Currently, some VMD decomposition parameters are affected by human experience factors and have a certain degree of subjectivity; some optimize the VMD decomposition, but there are defects in poor convergence and long calculation time during the iterative process, resulting in distortion of the reconstructed signal. The crested porcupine optimization (CPO) algorithm simulates the defensive behavior of porcupines. Through different protection mechanisms, it simulates the exploration and defense behaviors of porcupines. The algorithm introduces a cyclic population reduction technique to maintain the diversity of the population and accelerate the convergence speed. Only the porcupines that are truly threatened will activate the defense mechanism. During the algorithm optimization process, some crested porcupines are extracted from the population and reintroduced into the population to increase diversity and avoid falling into local minima. The empirical wavelet transform (EWT) analyzes signals from the time-frequency domain in multiple scales, dynamically generates wavelet bases, adaptively generates a low-pass filter and multiple band-pass filters based on the wavelet bases, and then spreads the signals in different frequency bands. It has high resolution and can capture the detailed changes of signals more precisely. Further, it calculates the signal similarity and eliminates small-correlation components, and can effectively separate signals from noise.

[0147] The noise reduction method includes the following steps:

[0148] Step S1: Obtain the true partial discharge signal of the power transformer body.

[0149] Specifically, obtaining the true partial discharge signal of the power transformer body specifically includes the following steps:

[0150] Step S11: Construct a partial discharge model; and build a partial discharge experimental platform;

[0151] Among them, the partial discharge experimental platform is a comprehensive signal acquisition and monitoring system. By connecting the signal acquisition system, sensors, oscilloscopes, and partial discharge detectors to form a complete network, it is used to monitor and analyze partial discharge phenomena;

[0152] Step S12: Based on the partial discharge experimental platform, adjust the voltage regulator to trigger the partial discharge model to generate a partial discharge signal, and observe the signal pattern through an oscilloscope and a partial discharge detector;

[0153] Step S13: Based on the signal pattern, collect and save the partial discharge pulse signal.

[0154] Step S2: Based on the target decomposition parameters, perform variational mode decomposition on the real partial discharge signal to obtain a number of intrinsic mode components. The target decomposition parameters are obtained by globally optimizing the variational mode decomposition parameters with the minimum value of the composite index of the residual index and the mutual information entropy as the fitness function global optimization objective and using the crown porcupine optimization algorithm.

[0155] Specifically, in step S2, the global optimization of the variational mode decomposition parameters using the crown porcupine optimization algorithm is obtained according to the defense mechanism;

[0156] The defense mechanism specifically includes the first defense mechanism, and / or the second defense mechanism, and / or the third defense mechanism, and / or the fourth defense mechanism;

[0157] The first defense mechanism is visual defense, and its strategy model is as follows:

[0158]

[0159] Where, represents the position of the i-th predator in the (t + 1)-th generation of the first defense mechanism; represents the optimal solution in the t-th generation of the first defense mechanism; represents the position of the i-th predator in the t-th generation of the first defense mechanism; represents the distance between the current porcupine and a randomly selected porcupine in the population; τ 1 , τ 2 represents a normally distributed random number, and the interval is [0, 1];

[0160] The second defense mechanism is sound defense, and its strategy model is as follows:

[0161]

[0162] Where, r 1 , r 2 represent random integers in [1, N]; τ 3 represents a random number between [0, 1]; U 2,1 represents a random probability between [0, 1], represents the position of the i-th predator in the (t + 1)-th generation of the second defense mechanism; represents the position of the i-th predator in the t-th generation of the second defense mechanism; y represents the distance between the current porcupine and a randomly selected porcupine in the population; represents the position of the predator when the random number is r 1 in the t-th generation of the second defense mechanism; represents the position of the predator when the random number is r 2 in the t-th generation of the second defense mechanism;

[0163] The third defense mechanism is that the crested porcupine defends itself by secreting stinking gases, and its strategy model is as follows:

[0164]

[0165] Among them,

[0166]

[0167] Among them, r 3 、r 4 、r 5 represent random integers in [1, N]; τ 3 represents a random number between [0, 1]; U 3,1 represents a random probability between [0, 1], represents the position of the i-th predator in the (t + 1)-th generation of the third defense mechanism; represents the position of the i-th predator in the t-th generation of the third defense mechanism; represents the position of the predator with a random number r 3 in the t-th generation of the third defense mechanism; represents the position of the predator with a random number r 4 in the t-th generation of the third defense mechanism; represents the fitness function of the i-th predator in the t-th generation of the third defense mechanism; τ 3 represents a normally distributed random number in the interval [0, 1]; δ 3 represents a random number in the interval [0, 1]; γ 3,t represents a time-related factor;

[0168] The fourth defense mechanism is that the crested porcupine attacks with its own quills, and its strategy model is as follows:

[0169]

[0170] Among them,

[0171]

[0172] Among them, τ 4 、τ 5 represent random numbers between [0, 1]; represents the position of the i-th predator in the (t + 1)-th generation of the fourth defense mechanism; represents the optimal solution in the t-th generation of the fourth defense mechanism; represents the position of the i-th predator in the t-th generation of the fourth defense mechanism; δ 4 represents a random number in the interval [0, 1]; γ 4,t represents a time-related factor; F i tDenote the fitness factor of the \(i\)-th predator in the \(t\)-th generation of the fourth defense mechanism; \(\alpha(1 - \tau 4 )+\tau 4 Denote the dynamic adjustment strategy, where \(\alpha\) is a control parameter with an interval in \([0, 1]\).

[0173] Establish an improved cyclic population reduction mechanism in the defense mechanism;

[0174] The model formula of the improved cyclic population reduction mechanism is as follows:

[0175]

[0176] In the formula, \(N 1 Denotes the population size; \(N min Denotes the minimum population; \(N'\) denotes the surrogate population size; \(t\) denotes the number of iterations; \(\%\) denotes the modulo operation; \(T max Denotes the maximum number of iterations; \(T\) denotes the number of cycles.

[0177] In step S2, the target decomposition parameter takes the minimum value of the composite index of the residual index and the mutual information entropy as the global optimization target of the fitness function, and the process of obtaining the fitness function is as follows:

[0178] Step S21: Calculate the residual index index, and the formula is:

[0179]

[0180] where \(N\) represents the length of the observed data; \(y j Denotes the \(j\)-th actual observed value; \(y'\) j Denotes the \(j\)-th predicted value;

[0181] Step S22: Calculate the mutual information entropy index, and the formula is:

[0182]

[0183] where \(p(x, y)\) represents the joint probability distribution of \(X\) and \(Y\); \(p(x)\) and \(p(y)\) represent the marginal probability distributions of \(X\) and \(Y\), \(X\) represents the original signal, and \(Y\) represents the intrinsic mode component obtained after decomposition;

[0184] In specific implementation, components with an absolute value of the correlation coefficient lower than 0.95 are discarded, and components that meet the threshold requirements are retained.

[0185] Step S23: Combine the residual index index and the mutual information entropy index by weighting to obtain a composite index;

[0186] The calculation formula of the composite index \(CI\) is;

[0187] \(CI=\lambda\times RI + \beta\times I\);

[0188] λ + β = 1;

[0189] Wherein, RI is the residual index, and I is the mutual information entropy index; both λ and β are penalty coefficients, which are positive numbers respectively and λ + β = 1.

[0190] Step S3: Further decompose the intrinsic mode components by using the EWT algorithm to obtain band mode components.

[0191] Step S4: Calculate the Pearson correlation coefficient between the band mode components and the intrinsic mode components; and discard the band mode components with a Pearson correlation coefficient lower than a preset threshold, while retaining the band mode components with a Pearson correlation coefficient not lower than the preset threshold.

[0192] Specifically, the Pearson correlation coefficient has the following calculation formula:

[0193]

[0194] Wherein, n represents the number of samples, and n is a natural number greater than 1; pccj ,k represents the correlation coefficient between the j-th component of IMFs and the k-th component of its corresponding EWT decomposition EWT_IMFs; IMFs j,i represents the i-th element of the j-th component of IMFs; EWT_IMFs j,k,i represents IMFs j,i the i-th element of the k-th component of the j-th component.

[0195] Step S5: Reconstruct the retained band mode components to obtain the denoised true partial discharge signal, thus completing the denoising of the multi-scale partial discharge signal based on CPO-optimized VMD_EWT.

[0196] Step S6: Verify the denoising effect through the partial discharge simulation signal;

[0197] The step S6 specifically includes the following steps:

[0198] Obtain the power of the partial discharge simulation signal before and after denoising respectively;

[0199] Calculate the signal-to-noise ratio SNR between the power of the partial discharge simulation signal before and after denoising sim ;

[0200]

[0201] Wherein,

[0202] Pow before is the power of the partial discharge simulation signal before denoising;

[0203] Powafter is the power of the partial discharge simulation signal after noise reduction;

[0204] Evaluate the noise reduction effect according to the signal-to-noise ratio.

[0205] As a specific implementation, the steps for obtaining the power of the partial discharge simulation signal are as follows:

[0206] Step A1: Obtain the pulse waveform of the partial discharge simulation signal;

[0207] The calculation formula for the pulse waveform of the partial discharge simulation signal is as follows:

[0208]

[0209] where s(t) represents the pulse waveform; A represents the maximum pulse amplitude; Ω1 represents the first pulse attenuation coefficient (which can take a value of 0.00005 in this embodiment); Ω2 represents the second pulse attenuation coefficient (which can take a value of 0.0000009 in this embodiment); t represents time.

[0210] Step A2: Calculate the reference power of the partial discharge simulation signal according to the pulse waveform of the partial discharge simulation signal, and its calculation formula is as follows:

[0211]

[0212] where s i represents the i-th point of the signal; N represents the signal length; sig_pow represents the signal power;

[0213] Step A3: Add Gaussian noise power to the reference power to obtain the power of the partial discharge simulation signal;

[0214] The calculation formula for the power of the partial discharge simulation signal is as follows:

[0215] P total = P base + P noise ;

[0216] where P toyal represents the power of the partial discharge simulation signal; P noise represents the Gaussian noise power;

[0217] The calculation formula for the Gaussian noise power is as follows:

[0218]

[0219] where randn is the standard Gaussian white noise; length is a function that returns the length of the array; SNR inputis the preset signal-to-noise ratio of Gaussian noise.

[0220] As a variation of this embodiment, as Figure 2 shown, the noise reduction method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT implemented in this embodiment includes the following steps: First, obtain the partial discharge simulation signal and the real signal. The simulated partial discharge pulse signal is composed of two decaying exponential functions. Different pulse durations determine the pulse width, and Gaussian white noise with a specified signal-to-noise ratio is added to the simulated pulse signal. In addition, the process of obtaining the partial discharge pulse signal of the real power transformer body includes establishing a typical partial discharge model structure, building a partial discharge experimental platform, installing and connecting signal acquisition devices, sensors, oscilloscopes, and partial discharge detectors and other equipment, adjusting the voltage regulator to trigger the partial discharge model, generating a partial discharge signal, and observing and recording the signal pattern and discharge inception voltage through the oscilloscope and partial discharge detector, and finally saving the partial discharge pulse signal. Next, perform modal decomposition on the ultra-high frequency (UHF) and high-frequency current transformer (HFCT) signals of partial discharge using the CPO_VMD algorithm, and use the minimum value of the composite index of the residual index and mutual information entropy as the fitness function for global optimization to obtain the optimal decomposition parameters. Then, the VMD algorithm decomposes the signal into several intrinsic mode functions and a residual component according to the optimized optimal decomposition parameters. Subsequently, apply the EWT algorithm to further decompose each intrinsic mode component and calculate its Pearson correlation coefficient. Discard the components with a correlation coefficient lower than the set threshold, and retain the modes with a correlation coefficient exceeding the set threshold. Finally, reconstruct the retained mode components to achieve signal noise reduction. The CPO optimization of VMD process includes various defense mechanisms, such as visual defense, sound defense, malodorous gas defense, and feather attack defense, ensures population diversity through the cyclic population reduction mechanism, and establishes a composite fitness function composed of the residual index and mutual information entropy to optimize the decomposition layer number and penalty factor, and saves the optimal decomposition parameters for subsequent use. Apply this algorithm to verify the simulated partial discharge pulse signal. By calculating the signal-to-noise ratio before and after noise reduction, the applicability and robustness of the algorithm are verified; at the same time, perform noise reduction processing on the partial discharge signal of the real transformer body to obtain several intrinsic mode components (IMFs) and a residual component (Rse). Further decompose each IMF through the EWT algorithm and calculate its Pearson correlation coefficient with the corresponding EWT decomposition component, filter out the components with a correlation coefficient lower than the set threshold, retain the mode components that meet the conditions and reconstruct them, and finally complete the noise reduction processing of the partial discharge signal, improve the signal-to-noise ratio, and achieve the purpose of effectively reducing the partial discharge signal of the power transformer body. (Specific formulas are omitted because they are the same as the previous calculation formulas). Figure 2 shows the flow diagram of the noise reduction method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT in the embodiment of the present invention, specifically including the following steps:

[0221] K0, Acquisition of partial discharge simulation signals and real signals.

[0222] In this step, first, a simulated partial discharge pulse signal is generated. This signal consists of two decaying exponential functions, and different pulse durations determine the pulse width. In addition, to simulate the noise environment in actual applications, Gaussian white noise with a specified signal-to-noise ratio is added to the simulated pulse signal. At the same time, the partial discharge pulse signal of the real power transformer body is acquired. The specific process includes establishing a typical partial discharge model structure, building a partial discharge experimental platform, installing and connecting signal acquisition devices, sensors, oscilloscopes, and partial discharge detectors and other equipment. By adjusting the voltage regulator to trigger the partial discharge model, an actual partial discharge signal is generated, and an oscilloscope and a partial discharge detector are used to observe and record the signal pattern and the discharge inception voltage, and finally, this partial discharge pulse signal is saved.

[0223] K1, Modal decomposition of partial discharge ultra-high frequency (UHF) and high-frequency current transformer (HFCT) signals using the CPO_VMD algorithm, with the minimum value of the composite index of the residual index and mutual information entropy as the fitness function for global optimization to obtain the optimal decomposition parameters.

[0224] In this step, for the ultra-high frequency (UHF) and high-frequency current transformer (HFCT) signals of partial discharge, the Crown Porcupine Optimization (CPO) algorithm combined with Variational Mode Decomposition (VMD) is applied for modal decomposition. The CPO algorithm simulates the multiple defense mechanisms of the crown porcupine to ensure that the population individuals can comprehensively explore the solution space during the iteration process and avoid falling into local optimal solutions. The optimization goal of modal decomposition is to minimize the two composite indices of the residual index and mutual information entropy, which are used as the fitness function for global optimization. Through this optimization process, the optimal decomposition parameters of the VMD algorithm, including the decomposition layer number and the penalty factor, are determined to provide the optimal parameter configuration for the subsequent signal decomposition steps.

[0225] K2, The VMD algorithm decomposes the signal using the optimal decomposition state to obtain several intrinsic mode functions and a residual component.

[0226] After obtaining the optimal decomposition parameters, the VMD algorithm is applied to decompose the input partial discharge signal. The VMD algorithm decomposes the signal into several intrinsic mode functions (IMFs) and a residual component according to the optimized decomposition state. These intrinsic mode functions represent different components of the signal at different center frequencies and bandwidths, reflecting the characteristics of the signal at different scales. The residual component contains the part of the signal that cannot be fully represented by the IMFs and is usually used to capture the low-frequency components in the signal or the noise that has not been decomposed.

[0227] K3. The EWT algorithm further decomposes each modal component, calculates the Pearson correlation coefficient, discards the components with a correlation coefficient less than the set threshold, and retains the modes with a correlation coefficient exceeding the set threshold.

[0228] Based on the VMD decomposition, the empirical wavelet transform (EWT) is further applied to subdivide each intrinsic mode component. The EWT algorithm adaptively constructs a wavelet filter bank based on the frequency characteristics of the signal itself to perform a more refined decomposition of each modal component. During this process, the Pearson correlation coefficient between each component and its corresponding EWT decomposition component is calculated. If the correlation coefficient of a certain component is lower than the preset threshold, it is considered that this component mainly contains noise or irrelevant signal components and should be discarded. Conversely, the modes with a correlation coefficient exceeding the threshold are considered to contain useful signal features and are retained. This process helps to further filter out noise and retain the components valuable for signal denoising and feature extraction.

[0229] K4. Reconstruct the retained modal components to finally achieve signal denoising.

[0230] After filtering out the irrelevant or noisy modal components, the remaining retained modal components are reconstructed. By recombining these retained modal components, a denoised signal is formed. This reconstruction process ensures that the useful partial discharge features in the signal are retained while the noise components are effectively removed, thus achieving the purpose of signal denoising. Finally, the denoised signal has a higher signal-to-noise ratio, facilitating subsequent fault diagnosis and monitoring analysis.

[0231] In summary, Figure 2 The whole process of the multi-scale partial discharge signal denoising method based on CPO optimized VMD_EWT is described in detail. From signal acquisition, optimizing modal decomposition parameters, signal decomposition, noise filtering to the final signal reconstruction, each step aims to improve the denoising effect and the accuracy of signal processing, ensuring that the true features of partial discharge signals can still be effectively extracted and analyzed in a complex electromagnetic environment.

[0232] In this embodiment, multi-scale partial discharge signal denoising is achieved through CPO optimized VMD_EWT, which can effectively avoid the local optimal convergence problem that may occur when the algorithm seeks the optimal solution, ensuring that the population individuals can explore the solution space more comprehensively during the iteration process, thereby improving the overall accuracy of the partial discharge denoising method. The specific process and effect are as Figures 3 to 15 .

[0233] As Figure 3 shown, Figure 3shows the detailed process schematic diagram in the embodiments of the present invention. The flowchart details the entire noise reduction process starting from initializing the CPO parameters, to VMD decomposition, calculating mutual information entropy and residual index, updating indicators, searching for agents, exploration phase, attack phase, updating positions, boundary constraints, until reaching the maximum number of iterations or convergence rate. As Figure 4 shown, Figure 4 is the time-domain diagram of the simulated partial discharge pulse signal. This diagram shows the waveform of the generated partial discharge pulse signal, which is used as the input for signal processing and displays the original form of the signal without added noise. As Figure 5 shown, Figure 5 is the time-domain diagram of the simulated partial discharge pulse signal with a signal-to-noise ratio of 10. This diagram shows the waveform of the signal after adding Gaussian white noise with a signal-to-noise ratio of 10 to the simulated pulse signal, and the influence of the noise on the signal can be clearly seen. As Figure 6 shown, Figure 6 shows the time-domain diagram of the simulated partial discharge pulse noisy signal after conventional EMD noise reduction. After noise reduction by the conventional empirical mode decomposition (EMD) method, part of the noise is removed, but the processing effect and clarity of the signal are still limited. As Figure 7 shown, Figure 7 shows the time-domain diagram of the simulated partial discharge pulse noisy signal after conventional EWT noise reduction. After noise reduction using the empirical wavelet transform (EWT) method, the noise components in the signal are further suppressed, and the noise reduction effect is improved compared to EMD. As Figure 8 shown, Figure 8 is the time-domain diagram of each mode of the simulated partial discharge pulse noisy signal after CPO_VMD decomposition. This diagram shows several intrinsic mode functions (IMFs) obtained by decomposing the noisy signal using the CPO-optimized VMD algorithm, and each mode reflects the main components of the signal at different frequencies and scales. As Figure 9 shown, Figure 9 shows the time-domain diagram of the simulated partial discharge pulse noisy signal after CPO_VMD noise reduction. After processing by the CPO-optimized VMD algorithm, the noise part of the signal is effectively suppressed, and the clarity of the signal is significantly improved. As Figure 10 shown, Figure 10 is the time-domain diagram of the simulated partial discharge pulse noisy signal after CPO_VMD and EWT noise reduction. Combining the double noise reduction of CPO-optimized VMD and EWT, the noise in the signal is almost completely removed, and the noise reduction effect is very significant. As Figure 11 shown, Figure 11 is the time-domain diagram of the floating discharge pulse signal. This diagram shows the waveform of the real floating discharge pulse signal, which is used for actual signal processing and analysis and reflects the discharge characteristics during equipment operation. As Figure 12 shown, Figure 12It shows the time-domain diagram of the floating discharge pulse signal after noise reduction by EWT. After the floating discharge pulse signal is processed by the EWT method for noise reduction, the noise components in the signal are weakened, but the noise reduction effect has not reached the best state. As Figure 13 shown, Figure 13 It shows the time-domain diagrams of each mode of the floating discharge pulse signal after decomposition by CPO_VMD. After the floating discharge pulse signal is decomposed by the VMD algorithm optimized by CPO, multiple intrinsic mode functions (IMFs) are obtained, and each mode represents the characteristics of the signal under different frequency components. As Figure 14 shown, Figure 14 It shows the time-domain diagram of the floating discharge pulse signal after noise reduction by CPO_VMD and EWT. After the floating discharge pulse signal is processed by combining CPO_VMD and EWT methods for noise reduction, the noise components in the signal are significantly suppressed, and the signal is restored more clearly. As Figure 15 shown, Figure 15 It shows the fitness curve diagram of the CPO_VMD iterative process. This diagram presents the change of the fitness value during the CPO optimization iteration process, showing how the algorithm is gradually optimized and finally converges to the optimal solution during the iteration process. To sum up, Figures 3 to 15 It details each processing step and its effect of the multi-scale partial discharge signal noise reduction method based on CPO-optimized VMD_EWT in the embodiment of the present invention, clearly reflecting the noise reduction ability and optimization effect of this method in different signal processing stages.

[0234] In this embodiment, an innovative multi-scale partial discharge signal noise reduction method is proposed, which optimizes the variational mode decomposition (VMD) based on the crown porcupine optimization algorithm (CPO). This method is specifically designed to solve the problems that may occur when traditional optimization algorithms are combined with VMD, such as being prone to falling into local optimal solutions, slow decomposition speed, and low decomposition accuracy. By optimizing VMD with CPO, it can conduct a comprehensive search in the parameter space, effectively avoid the trap of local optimal solutions, and accelerate the decomposition speed and improve the decomposition accuracy of the algorithm.

[0235] During the VMD decomposition process, a composite index composed of mutual information entropy and residual index is used as the fitness function. Compared with traditional single indexes, this composite index can consider information more comprehensively, thereby reducing the phenomena of mode mixing and mode leakage. By dynamically adjusting the weights of the residual index and mutual information entropy, the best combination of decomposition parameters can be obtained. Among them, the residual index is used to measure the thoroughness of decomposition to avoid mode leakage; the mutual information entropy is used to measure the correlation of mode components to avoid mode mixing phenomena.

[0236] In addition, this embodiment also introduces the empirical wavelet transform (EWT) for secondary decomposition to further analyze the signal details. EWT can extract the detailed features and local features in the signal, better capture the instantaneous changes and mutations of the signal. Especially for partial discharge pulse signals, it can separate more refined modes. Under the limitation of the Pearson correlation coefficient threshold, this method can improve the quality and reliability of the decomposition results and further improve the signal-to-noise ratio.

[0237] To sum up, this embodiment first uses the residual index and mutual information entropy as fitness functions to optimize the variational mode decomposition, and combines the multi-scale signal denoising method to further perform secondary decomposition on the basis of CPO-optimized VMD. This method not only improves the quality of the decomposition results, but also significantly helps in processing the noise of partial discharge signals, can effectively identify and remove the noise in different frequency bands, thereby improving the signal-to-noise ratio of the signal.

[0238] Embodiment 2:

[0239] This embodiment provides a method for detecting partial discharge signals, and the method includes the following steps:

[0240] Adopt the noise reduction method for multi-scale partial discharge signals based on CPO-optimized VMD_EWT described in Embodiment 1 to obtain the true partial discharge signal after noise reduction;

[0241] Detect the true partial discharge signal after noise reduction.

[0242] Embodiment 3:

[0243] As Figure 16 shown, this embodiment provides a noise reduction device for multi-scale partial discharge signals based on CPO-optimized VMD_EWT, and the device includes:

[0244] An acquisition unit 10, configured to acquire the true partial discharge signal of the power transformer body;

[0245] A first decomposition unit 20, connected to the acquisition unit 10, configured to perform variational mode decomposition on the true partial discharge signal based on target decomposition parameters to obtain a plurality of intrinsic mode components;

[0246] Wherein, the target decomposition parameters are obtained by taking the minimum value of the composite index of the residual index and mutual information entropy as the global optimization target of the fitness function, and using the crown porcupine optimization algorithm to globally optimize the variational mode decomposition parameters;

[0247] A second decomposition unit 30, connected to the first decomposition unit 20, configured to further decompose the intrinsic mode components by using the EWT algorithm to obtain frequency band mode components;

[0248] The selection unit 40, which is connected to the second decomposition unit 30, is used to discard the band modal components with Pearson correlation coefficients lower than a preset threshold, while retaining the band modal components with Pearson correlation coefficients not lower than the preset threshold;

[0249] The reconstruction unit 50, which is connected to the selection unit 40, is used to reconstruct the retained band modal components to obtain the denoised true partial discharge signal, thereby completing the denoising of the multi-scale partial discharge signal based on CPO-optimized VMD_EWT.

[0250] As a specific implementation, the device further includes an inspection unit;

[0251] The inspection unit is used to inspect the denoising effect through the partial discharge simulation signal;

[0252] The inspection unit includes:

[0253] An acquisition module, which is used to acquire the powers of the partial discharge simulation signals before and after denoising respectively;

[0254] A calculation module, which is connected to the acquisition module, is used to calculate the signal-to-noise ratio SNR between the powers of the partial discharge simulation signals before and after denoising sim ;

[0255] The calculation formula of the signal-to-noise ratio SNR is stored in the calculation module sim as follows:

[0256]

[0257] where

[0258] Pow before is the power of the partial discharge simulation signal before denoising;

[0259] Pow after is the power of the partial discharge simulation signal after denoising;

[0260] An evaluation module, which is connected to the calculation module, is used to evaluate the denoising effect according to the signal-to-noise ratio.

[0261] The device in this embodiment can execute the method in Embodiment 1.

[0262] Embodiment 4:

[0263] This embodiment provides a power transformer system, which includes a memory and a processor. A computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the denoising method of the multi-scale partial discharge signal based on CPO-optimized VMD_EWT described in Embodiment 1, or executes the detection method described in the second aspect.

[0264] It is understood that the above embodiments are merely exemplary embodiments adopted for the purpose of illustrating the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also regarded as the protection scope of the present invention.

Claims

1. A denoising method for multi-scale partial discharge signals of VMD_EWT based on CPO optimization, characterized in that: The method comprises the following steps: Step S1: obtaining a real partial discharge signal of a power transformer body; Step S2: performing variational mode decomposition on the real partial discharge signal based on the target decomposition parameter to obtain a number of intrinsic mode components; The target decomposition parameters are obtained by taking the minimum value of the composite index of residual index and mutual information entropy as the global optimization target of the fitness function, and using the crown porcupine optimization algorithm to perform global optimization on the variational mode decomposition parameters; Step S3: using the EWT algorithm to further decompose the intrinsic mode components to obtain frequency band mode components; Step S4: calculating the Pearson correlation coefficient between the band modal component and the eigenmodal component; and, Discarding frequency band modal components whose Pearson correlation coefficient is lower than a preset threshold, while retaining frequency band modal components whose Pearson correlation coefficient is not lower than the preset threshold; Step S5: Reconstruct the retained frequency band modal components to obtain the real partial discharge signal after denoising, thereby completing the denoising of multi-scale partial discharge signal based on CPO optimized VMD_EWT.

2. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 1 is characterized in that: In the step S2, the variational modal decomposition parameters are globally optimized using the crested porcupine optimization algorithm, which is obtained based on the defense mechanism; The defense mechanism specifically includes a first defense mechanism, and / or a second defense mechanism, and / or a third defense mechanism, and / or a fourth defense mechanism; The first defense mechanism is visual defense, and its strategy model is as follows: in, represents the position of the i-th predator in the t+1th generation of the first defense mechanism; represents the optimal solution of the tth generation of the first defense mechanism; represents the position of the i-th predator in the t-th generation of the first defense mechanism; represents the distance between the current porcupine and a random porcupine in the population; τ1, τ2 represent normally distributed random numbers, the interval is [0,1]; The second defense mechanism is sound defense, and its strategy model is as follows: Among them, r1 and r2 represent random integers between [1, N]; τ3 represents a random number between [0, 1]; U 2,1 represents the random probability between [0,1], represents the position of the i-th predator in the t+1th generation of the second defense mechanism; represents the position of the i-th predator in the t-th generation of the second defense mechanism; y represents the distance between the current porcupine and a random porcupine in the population; It indicates the position of the predator with random number r1 in the tth generation of the second defense mechanism; It indicates the random number r2 predator position in the tth generation of the second defense mechanism; The third defense mechanism is that the crested porcupine defends by secreting foul-smelling gas, and its strategy model is as follows: Among them, r3, r4, r5 represent random integers between [1, N]; τ3 represents a random number between [0, 1]; U 3,1 represents the random probability between [0,1], represents the position of the i-th predator in the t+1th generation of the third defense mechanism; represents the position of the i-th predator in the t-th generation of the third defense mechanism; It indicates the random number r3 predator position in the tth generation of the third defense mechanism; It indicates the position of the predator with random number r4 in the tth generation of the third defense mechanism; represents the fitness function of the tth generation i of the third defense mechanism; τ3 represents a normally distributed random number in the interval [0,1]; δ3 represents a random number in the interval [0,1]; γ 3,t represents the time-related factor; The fourth defense mechanism is that the crested porcupine uses its own feathers to attack, and its strategy model is as follows: Among them, τ4 and τ5 represent random numbers between [0,1]; represents the position of the i-th predator in the t+1th generation of the fourth defense mechanism; represents the t-th generation optimal solution of the fourth defense mechanism; represents the position of the i-th predator in the t-th generation of the fourth defense mechanism; δ4 represents a random number in the interval [0,1]; γ 4,t represents the time-related factor; represents the fitness factor of the i-th predator of the t-th generation of the fourth defense mechanism; α(1-τ4)+τ4 represents the dynamic adjustment strategy, α is the control parameter, and the interval is [0,1].

3. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 2 is characterized in that: Establishing improved cyclic population reduction mechanisms within said defense mechanisms; The model formula of the improved cyclic population reduction mechanism is as follows: In the formula, N1 represents the population size; % represents the remainder operation; N min represents the minimum population; N′ represents the number of agent populations; t represents the number of iterations; T max represents the maximum number of iterations; T represents the number of loops.

4. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 1, characterized in that: In step S2, the target decomposition parameter is a global optimization target of the fitness function with the minimum value of the residual index and the mutual information entropy composite index, wherein the fitness function is obtained as follows: Step S21: Calculate the residual index indicator, the formula is: Where N represents the length of the observation data; y j represents the jth actual observation value; y′ j represents the jth predicted value; Step S22: Calculate the mutual information entropy index, the formula is: Where p(x,y) represents the joint probability distribution of X and Y; p(x) and p(y) represent the marginal probability distribution of X and Y, X represents the original signal, and Y represents the intrinsic mode component obtained after decomposition; Step S23: performing weighted combination of the residual index indicator and the mutual information entropy indicator to obtain a composite indicator; The calculation formula of the composite index CI is: CI = λ × RI + β × I; λ+β=1; Among them, RI is the residual index indicator, I is the mutual information entropy indicator; λ and β are penalty coefficients, which are positive numbers and λ+β=1.

5. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 1, characterized in that: The calculation formula of the Pearson correlation coefficient in step S4 is as follows: Where n represents the number of samples, and n is a natural number greater than 1; pcc j,k represents the correlation coefficient between the jth component of IMFs and its corresponding EWT decomposition EWT_IMFs kth component; IMFs j,i Represented as the i-th element of the j-th component of IMFs; EWT_IMFs j,k,i Represented as IMFs j,i The i-th element of the k-th component of the j-th component.

6. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 1, characterized in that: The real partial discharge signal is a partial discharge ultra-high frequency signal and / or a pulse current signal; In the step S2, performing variational mode decomposition on the real partial discharge signal specifically includes: The partial discharge ultra-high frequency signal is subjected to variational modal decomposition, and / or the pulse current signal is subjected to variational modal decomposition.

7. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 1, characterized in that: The step S1 specifically includes the following steps: Step S11: constructing a partial discharge model; and building a partial discharge experimental platform; Step S12: Based on the partial discharge experimental platform, adjusting the voltage regulator to trigger the partial discharge model to generate a partial discharge signal, and observing the signal spectrum through an oscilloscope and a partial discharge detector; Step S13: Based on the signal spectrum, the partial discharge pulse signal is collected and saved to complete the acquisition of the real partial discharge signal of the power transformer body.

8. The denoising method for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to any one of claims 1 to 7, characterized in that: After step S5, the method further includes: Step S6: testing the noise reduction effect through partial discharge simulation signal; The step S6 specifically includes the following steps: Obtain the power of the partial discharge simulation signal before and after noise reduction respectively; Calculate the signal-to-noise ratio (SNR) between the power of the partial discharge simulation signal before and after noise reduction sim ; in, Pow before is the power of the partial discharge simulation signal before noise reduction; Pow after is the power of the partial discharge simulation signal after noise reduction; The noise reduction effect is evaluated according to the signal-to-noise ratio.

9. The method for denoising multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 8, characterized in that: The steps for obtaining the partial discharge simulation signal power are as follows: Step A1: obtaining a pulse waveform of a partial discharge simulation signal; The calculation formula of the pulse waveform of the partial discharge simulation signal is as follows: Where, s(t) represents the pulse waveform; A represents the maximum amplitude of the pulse; Ω1 represents the first pulse attenuation coefficient; Ω2 represents the second pulse attenuation coefficient; Step A2: Calculate the reference power P of the partial discharge simulation signal according to the pulse waveform of the partial discharge simulation signal base , and its calculation formula is as follows: Among them, s i represents the i-th point of the signal; N represents the length of the signal; Step A3: adding Gaussian noise power to the reference power to obtain partial discharge simulation signal power; The calculation formula of the partial discharge simulation signal power is as follows: P total =P base +P noise ; Among them, P total Represents the power of partial discharge simulation signal; P noise represents Gaussian noise power; P base Indicates the reference power; The calculation formula for Gaussian noise power is as follows: Among them, randn is standard Gaussian white noise; length is the function that returns the length of the array; SNR input is the preset Gaussian noise signal-to-noise ratio.

10. A method for detecting a partial discharge signal, characterized in that: The method comprises the following steps: Adopting the denoising method of multi-scale partial discharge signal based on CPO optimization of VMD_EWT as described in any one of claims 1 to 9, to obtain a real partial discharge signal after denoising; The real partial discharge signal after noise reduction is detected.

11. A noise reduction device for multi-scale partial discharge signals based on CPO optimization of VMD_EWT, characterized in that: include: An acquisition unit, used for acquiring a real partial discharge signal of a power transformer body; a first decomposition unit, connected to the acquisition unit, and configured to perform variational mode decomposition on the real partial discharge signal based on a target decomposition parameter to obtain a plurality of intrinsic mode components; The target decomposition parameters are obtained by taking the minimum value of the composite index of residual index and mutual information entropy as the global optimization target of the fitness function, and using the crown porcupine optimization algorithm to perform global optimization on the variational mode decomposition parameters; a second decomposition unit, connected to the first decomposition unit, and configured to further decompose the intrinsic mode component by using an EWT algorithm to obtain a frequency band mode component; a selection unit connected to the second decomposition unit, configured to discard the frequency band modal components whose Pearson correlation coefficient is lower than a preset threshold, and retain the frequency band modal components whose Pearson correlation coefficient is not lower than the preset threshold; The reconstruction unit is connected to the selection unit and is used to reconstruct the retained frequency band modal components to obtain the real partial discharge signal after noise reduction, thereby completing the denoising of the multi-scale partial discharge signal based on CPO optimization VMD_EWT.

12. The denoising device for multi-scale partial discharge signals based on CPO optimization of VMD_EWT according to claim 11, characterized in that: Also includes an inspection unit; The testing unit is used to test the noise reduction effect through a partial discharge simulation signal; The inspection unit comprises: An acquisition module, used to respectively acquire the power of the partial discharge simulation signal before and after noise reduction; A calculation module connected to the acquisition module is used to calculate the signal-to-noise ratio (SNR) between the power of the local discharge simulation signal before and after noise reduction. sim ; The calculation module stores the signal-to-noise ratio SNR sim The calculation formula is as follows: in, Pow before is the power of the partial discharge simulation signal before noise reduction; Pow after is the power of the partial discharge simulation signal after noise reduction; An evaluation module is connected to the calculation module and is used to evaluate the noise reduction effect according to the signal-to-noise ratio.

13. A power transformer system, characterized in that: It includes a memory and a processor, wherein a computer program is stored in the memory. When the processor runs the computer program stored in the memory, the processor executes the denoising method for multi-scale local discharge signals of VMD_EWT based on CPO optimization according to any one of claims 1 to 9, or the detection method according to claim 10.

Images