Self-adaptive extraction method, system and equipment for weak discharge signal of transformer and medium

Abstract

The invention belongs to the technical field of partial discharge detection, and particularly relates to a transformer weak discharge signal self-adaptive extraction method, system and device and a medium. In order to overcome the defect that the existing white noise removal effect is poor when a partial discharge signal is weak in the partial discharge detection process of an existing transformer, the invention adopts the following technical scheme: a transformer weak discharge signal self-adaptive extraction method under serious background noise interference is adopted. Comprising the following steps: combining a wavelet decomposition layer number, a threshold adjustment coefficient and a thresholding function adjustment coefficient into a parameter group, taking the parameter group as a particle, establishing a particle swarm, and initializing the particle; improving a threshold function and a thresholding function in a wavelet threshold denoising method based on each parameter in the particles, processing an input signal by adopting the improved functions, and traversing all the particles; and optimizing by adopting a particle swarm optimization algorithm to obtain an optimal parameter, and outputting a denoising result. The method provided by the invention has a good denoising effect on weak discharge signals under serious background noise interference.

Description

Technical Field

[0001] The invention belongs to the technical field of partial discharge detection, and in particular relates to a method, system, equipment and medium for adaptively extracting transformer weak discharge signals. Background Art

[0002] As the core equipment of the power transmission system, the operating status of the converter transformer directly affects the safe and stable operation of the entire system. Partial discharge can cause aging and deterioration of insulation materials, which is an important cause of equipment insulation failure and may cause serious safety accidents. By accurately detecting the partial discharge of the converter transformer, the insulation condition of the converter transformer can be effectively obtained, and potential defects in the converter transformer can be discovered in time, which is of great significance to ensure the safety of the equipment and the reliable operation of the system.

[0003] Ultrasonic emission measurement is an important technology for detecting and locating partial discharge in power transformers. It plays an important role in evaluating the insulation status of transformers and ensuring the safe and stable operation of power systems. However, due to the large size and complex internal structure of large transformers such as converter transformers, the ultrasonic signal generated by the discharge will experience significant energy attenuation during the propagation process, causing the weak discharge signal to be severely distorted or even submerged by white noise, which has a great negative impact on partial discharge monitoring and positioning. Therefore, accurately and completely extracting the original ultrasonic signal from the white noise and minimizing the interference is the key to the success of ultrasonic detection of partial discharge in transformers.

[0004] Wavelet threshold denoising is a commonly used method. After wavelet decomposition, the amplitude of the signal's wavelet coefficient is greater than the amplitude of the noise coefficient. It can be considered that the wavelet coefficient with a larger amplitude is generally dominated by the signal, while the coefficient with a smaller amplitude is largely noise. Therefore, the threshold method is used to reduce most of the noise coefficients to 0, that is, the coefficients smaller than the threshold are removed as interference noise, and then wavelet reconstruction is performed to achieve noise reduction.

[0005] In wavelet threshold denoising, the thresholding function reflects different processing strategies for wavelet coefficients above and below the threshold, and is a key step in threshold denoising. There are two traditional threshold functions, namely hard threshold and soft threshold. In view of the shortcomings of traditional thresholds, some improved wavelet threshold denoising methods have emerged. These methods are mainly based on the thresholding function. The core purpose of these methods is to eliminate the discontinuity of the threshold function while making the function quickly approach the hard threshold function.

[0006] The number of wavelet decomposition layers, thresholds, and thresholding functions are all important parameters that can significantly affect the denoising effect. However, these parameters are either not considered comprehensively, or the optimization method has room for improvement. In addition, how to accurately evaluate the denoising effect also affects the selection of these parameters. Researchers often need to spend a lot of time making adjustments to get a better denoising effect, which makes the denoising algorithm not convenient enough in field applications. Therefore, it is necessary to improve the existing algorithms to improve their reliability and ease of use in practical applications. Summary of the invention

[0007] The present invention aims at the insufficiency of the existing white noise removal effect when the partial discharge signal is relatively weak in the existing transformer partial discharge detection process, and provides a transformer weak discharge signal adaptive extraction method under severe background noise interference, improves the wavelet threshold and threshold function, and through a more accurate evaluation of the denoising effect, realizes the automatic optimization of relevant parameters in the denoising process, and finally improves the denoising effect. The present invention also provides a transformer weak discharge signal adaptive extraction system, a computer device and a readable storage medium.

[0008] To achieve the above object, the present invention adopts the following technical solution: a method for adaptively extracting weak discharge signals of transformers under severe background noise interference, the method for adaptively extracting weak discharge signals of transformers under severe background noise interference comprises:

[0009] Step S1, combining the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and treating it as a particle, randomly assigning initial values ​​to the positions of the three parameters in each particle within a value range, and at the same time, randomly assigning initial values ​​to the initial speed of each particle within a value range, randomly generating N particles, and establishing a particle swarm;

[0010] Step S2, based on the parameters in the particles, the threshold function and the thresholding function in the wavelet threshold denoising method are improved, the improved threshold function and the thresholding function are used to process the input signal A, and all particles are traversed to obtain N results;

[0011] Step S3, updating the moving speed and position of each particle;

[0012] Step S4, calculating the fitness value of each particle;

[0013] Step S5, updating the individual historical optimal solution of each particle according to the fitness value;

[0014] Step S6, updating the global optimal solution of the particle swarm according to the fitness value;

[0015] Step S7, repeat the process from step S3 to step S6 until the maximum number of iterations is reached or the global optimal solution of the particle no longer changes, and obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m at this time as the optimal solution, and process the input signal A accordingly to output the denoising result.

[0016] The method for adaptively extracting weak discharge signals of transformers under severe background noise interference of the present invention combines the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group as a particle, and adopts a particle swarm optimization algorithm to automatically optimize the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m. It can quickly obtain better parameters on site, has strong reliability and high practicality, and has a good denoising effect on weak discharge signals under severe background noise interference.

[0017] As an improvement, in step S2, the improved thresholding function is expressed as:

[0018]

[0019] Where, d s (j,L) is the function of the input signal A on the jth layer after L layers of wavelet decomposition, d' s (j,L) is the value of d s (j, L) is the output function after thresholding, T is the set threshold, and m is the custom thresholding function adjustment coefficient.

[0020] As an improvement, in step S2, the improved threshold function is expressed as:

[0021] T=c×max(d n (j,L))

[0022] Where c is the threshold adjustment coefficient d n (j,L) represents the noise segment A in the input signal A N The function on the jth layer after L-layer wavelet decomposition, A N The first 10% or last 10% of signal A.

[0023] As an improvement, in step S3, after t iterations, the three parameters L, c, and m of each particle are updated using the following formulas for the t+1th iteration:

[0024] v L (t+1)=w·v L (t)+c1·r1·(p L_best (t)-L(t))+c2·r2·(g L_best (t)-L(t))

[0025] v c (t+1)=w·v c (t)+c1·r1·(p c_best (t)-c(t))+c2·r2·(g c_best (t)-c(t))

[0026] v m (t+1)=w·v m (t)+c1·r1·(p m_best (t)-m(t))+c2·r2·(g m_best (t)-m(t))

[0027] In the formula, v L 、v c 、v m are the speeds of L, c, and m respectively, w is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers between [0,1], and p L_best 、p c_best 、p m_best are the individual historical optimal solutions of L, c, and m, respectively, and g L_best ,g c_best ,g m_best are the global optimal solutions of L, c, and m respectively.

[0028] As an improvement, in step S4, QNR is used as the fitness evaluation index of the particle, and the calculation formula of QNR is:

[0029]

[0030] Where l is the length of the signal, p is the number of delay cycles specified, is the delayed k-order autocorrelation coefficient of the noise residual. The smaller the QNR, the closer the noise residual is to white noise, indicating that the denoised signal is closer to the partial discharge signal.

[0031] As an improvement, in step S5, if the fitness value of the current particle is better than the optimal value of the fitness in the particle's history, the individual historical optimal solution of the particle is updated; in step S6, if the fitness value of the current particle is better than the global optimal value of the current fitness, the global optimal solution of the entire particle swarm is updated.

[0032] As an improvement, in step S1, the value range of the wavelet decomposition layer number L is 4 to 20, the value range of the threshold adjustment coefficient c is 0.2 to 10, and the value range of the threshold function adjustment coefficient m is 2 to 2000.

[0033] The transformer weak discharge signal adaptive extraction system under severe background noise interference comprises:

[0034] The particle swarm construction module is used to combine the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and use it as a particle. The positions of the three parameters in each particle are randomly assigned initial values ​​within the value range. At the same time, the initial speed of each particle is randomly assigned an initial value within the value range. N particles are randomly generated to establish a particle swarm.

[0035] The wavelet threshold denoising module is used to improve the threshold function and thresholding function in the wavelet threshold denoising method based on the parameters in the particles, use the improved threshold function and thresholding function to process the input signal A, traverse all particles, and obtain N results;

[0036] Particle update module, used to update the moving speed and position of each particle;

[0037] The fitness value calculation module is used to calculate the fitness value of each particle;

[0038] Individual historical optimal solution update module, used to update the individual historical optimal solution of each particle according to the fitness value;

[0039] The global optimal solution updating module is used to update the global optimal solution of the particle swarm according to the fitness value;

[0040] The output module is used to obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m as the optimal solution when the maximum number of iterations is reached or the global optimal solution of the particle no longer changes. The input signal A is processed accordingly and the denoising result is output.

[0041] The computer device includes a processor and a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by the processor, the method for adaptively extracting transformer weak discharge signals under severe background noise interference is implemented.

[0042] A computer-readable storage medium stores a computer program, which, when executed, implements the aforementioned method for adaptively extracting transformer weak discharge signals under severe background noise interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for adaptively extracting transformer weak discharge signals under severe background noise interference according to an embodiment of the present invention.

[0044] Figure 2 3 is a comparison diagram between the improved new thresholding function according to the embodiment of the present invention and the traditional thresholding function.

[0045] Figure 3 is the partial discharge signal generated by simulation.

[0046] Figure 4 It is the noisy partial discharge signal generated by simulation (including partial discharge signal and noise signal).

[0047] Figure 5 The method of the embodiment of the present invention is used to Figure 4 The resulting image after denoising.

[0048] Figure 6 It is the actual collected partial discharge ultrasonic signal of a transformer.

[0049] Figure 7 The method of the embodiment of the present invention is used to Figure 6 The resulting image after denoising.

[0050] Figure 8 An existing method is used to Figure 6 The resulting image after denoising. DETAILED DESCRIPTION

[0051] The technical solutions of the embodiments of the present invention are explained and described below, but the following embodiments are only preferred embodiments of the present invention, not all. Based on the embodiments in the implementation mode, other embodiments obtained by those skilled in the art without creative work are all within the protection scope of the present invention.

[0052] See also Figures 1 to 7 , the transformer weak discharge signal adaptive extraction method under severe background noise interference of an embodiment of the present invention, the transformer weak discharge signal adaptive extraction method under severe background noise interference comprises:

[0053] Step S1, combining the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and treating it as a particle, randomly assigning initial values ​​to the positions of the three parameters in each particle within a value range, and at the same time, randomly assigning initial values ​​to the initial speed of each particle within a value range, randomly generating N particles, and establishing a particle swarm;

[0054] Step S2, based on the parameters in the particles, the threshold function and the thresholding function in the wavelet threshold denoising method are improved, the improved threshold function and the thresholding function are used to process the input signal A, and all particles are traversed to obtain N results;

[0055] Step S3, updating the moving speed and position of each particle;

[0056] Step S4, calculating the fitness value of each particle;

[0057] Step S5, updating the individual historical optimal solution of each particle according to the fitness value;

[0058] Step S6, updating the global optimal solution of the particle swarm according to the fitness value;

[0059] Step S7, repeat the process from step S3 to step S6 until the maximum number of iterations is reached or the global optimal solution of the particle no longer changes, and obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m at this time as the optimal solution, and process the input signal A accordingly to output the denoising result.

[0060] In this embodiment, each particle represents a possible solution. The size of the particle swarm, i.e., the number of particles N, is usually in the range of 20 to 100, and is set to 50 in this embodiment. The positions of the three parameters in each particle are randomly assigned initial values ​​(L(0), c(0), m(0)) within the range; at the same time, the initial velocity of each particle (v L (0), v c (0), v m (0)), is randomly assigned an initial value within the value range.

[0061] In this embodiment, in step S2, the improved thresholding function is expressed as:

[0062]

[0063] Where, d s (j,L) is the function of the input signal A on the jth layer after L layers of wavelet decomposition, d' s (j,L) is the value of d s (j, L) is the output function after thresholding, T is the set threshold, and m is the custom thresholding function adjustment coefficient.

[0064] The improved thresholding function can be understood as s The part of (j,L) that is greater than the threshold T, output d' s (j,L) equals d s (j,L), for d s The part of (j,L) that is less than the threshold T, output d' s (j,L) will quickly approach 0 with a curve, thus realizing the thresholding function, that is, the part greater than the threshold T is retained, and the part less than the threshold T is set to zero. Compared with the traditional soft threshold function, the optimized function can eliminate the s The fixed deviation generated when (j,L) is greater than the threshold. Compared with the traditional hard threshold function, the optimized function can solve the instability problem at the threshold point.

[0065] The comparison between the optimized thresholding function and the traditional thresholding function is as follows: Figure 2 As shown. Figure 2 It can be seen that when the threshold is set to 5,s When (j,L) is 10, the output of the traditional soft threshold function is d' s (j, L) is 5, which is inconsistent with the output of 10. However, the function of this embodiment and the traditional hard threshold function can both output 10. s When (j, L) is 5, the output of the traditional hard threshold function corresponds to multiple output values ​​in the range of 0-5, resulting in unstable output, while the function proposed in this embodiment can stably output 5; when d s When (j, L) is less than 5, the output of the function in this embodiment will quickly approach 0.

[0066] In this embodiment, in step S2, the improved threshold function is expressed as:

[0067] T=c×max(d n (j,L))

[0068] Where c is the threshold adjustment coefficient, d n (j,L) represents the noise segment A in the input signal A N The function on the jth layer after L-layer wavelet decomposition, A N It is the first 10% or last 10% of signal A. Generally, the waveform on site will be pre-triggered, and the signal will decay quickly, so the first and last 10% are usually dominated by noise energy. Collecting the first and last 10% is in line with the actual situation on site.

[0069] In this embodiment, in step S3, after t iterations, the three parameters L, c, and m of each particle at the t+1th time are updated using the following formulas to update its speed and position respectively:

[0070] v L (t+1)=w·v L (t)+c1·r1·(p L_best (t)-L(t))+c2·r2·(g L_best (t)-L(t))

[0071] v c (t+1)=w·v c (t)+c1·r1·(p c_best (t)-c(t))+c2·r2·(g c_best (t)-c(t))

[0072] v m (t+1)=w·v m (t)+c1·r1·(p m_best (t)-m(t))+c2·r2·(g m_best (t)-m(t))

[0073] In the formula, vL 、v c 、v m are the speeds of L, c, and m respectively; w is the inertia weight, which controls the inertia of the particle, that is, its tendency to maintain the current speed. The typical value range is 0.4-0.9, and it is generally reduced from a larger value to achieve the transition from global search to local search during the iteration process. In this embodiment, it is selected to start from 0.9, and it is reduced by 0.1 every 5 iterations until it remains unchanged at 0.4; c1 and c2 are acceleration constants, which are used to adjust the speed of the particle moving to the individual extreme value and the global extreme value, and the value is usually c1=c2=2; r1 and r2 are random numbers between [0,1], which are used to introduce randomness and increase the diversity of the algorithm; p L_best 、p c_best 、p m_best are the individual historical optimal solutions of L, c, and m, respectively, and g L_best ,g c_best ,g m_best They are the global optimal solutions of L, c, and m respectively. The speed of the three parameters L, c, and m does not exceed the value range of the speed. When it is greater than the maximum value, the maximum value is taken, and when it is less than the minimum value, the minimum value is taken.

[0074] In this embodiment, in step S4, QNR is used as the fitness evaluation index of the particle, and the calculation formula of QNR is:

[0075]

[0076] Where l is the length of the signal, p is the number of delay cycles specified, is the delayed k-order autocorrelation coefficient of the noise residual. The smaller the QNR, the closer the noise residual is to white noise, indicating that the denoised signal is closer to the partial discharge signal. For each particle, calculate the fitness value QNR corresponding to its current position (L, c, m). By using the evaluation index QNR as the fitness function in the particle swarm optimization algorithm, the denoising effect can be evaluated each time. By constantly moving the position of the particle swarm to minimize QNR, the optimal particle can be obtained, that is, the optimal value of the wavelet decomposition layer number L, the threshold function adjustment coefficient m, and the threshold function adjustment coefficient c in the denoising process. This optimal value can achieve the best denoising effect.

[0077] In this embodiment, in step S5, if the fitness value of the current particle is better than the optimal value of the fitness in the particle's history, the individual historical optimal solution of the particle is updated; in step S6, if the fitness value of the current particle is better than the global optimal value of the current fitness, the global optimal solution of the entire particle swarm is updated.

[0078] In other embodiments, in step S1, the value range of the wavelet decomposition layer number L is 4-20, the value range of the threshold adjustment coefficient c is 0.2-10, and the value range of the threshold function adjustment coefficient m is 2-2000.

[0079] The main process of the adaptive extraction method of transformer weak discharge signal under severe background noise interference of the embodiment of the present invention includes: first, creating a particle group containing 50 particles, each particle includes three parameters: wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m, and taking initial values ​​for these three parameters; then, improving the threshold function based on the threshold function adjustment coefficient m and the threshold adjustment coefficient c, improving the threshold function based on the threshold function and the wavelet decomposition layer number L, and using the improved wavelet threshold denoising method to process the input signal to obtain denoising results under different particles (at least one of the three parameters is different); thirdly, updating the moving speed and position of each particle (that is, changing The size of at least one of the three parameters is calculated, the fitness value of each particle is calculated, the individual historical optimal solution of each particle is updated according to the fitness value, the global optimal solution of the particle swarm is updated according to the fitness value, and QNR is used as the fitness evaluation index of the particle; finally, when the maximum number of iterations is reached or the global optimal solution of the particle no longer changes, the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m at this time are obtained as the optimal solution (such as L is 10, c is 1, and m is 5. This optimal solution is likely not in the initial 50 particles). These three parameters are input into the wavelet threshold denoising method and the input signal A is processed to obtain the denoising result.

[0080] The method for adaptively extracting weak discharge signals of transformers under severe background noise interference of the embodiment of the present invention first generates some different parameter settings, which determine how to process the signal to remove noise. Each set of parameters will be used to process the signal, and then the denoising effect will be evaluated. Then, the parameters will be adjusted continuously according to these evaluation results to find the most suitable settings. After multiple rounds of optimization, the most effective parameter combination can be selected and used to process the signal, ensuring that the weak discharge signal can still be clearly extracted under the condition of large noise interference.

[0081] Although the particle swarm optimization algorithm (PSO) has been applied to partial discharge signal wavelet denoising in the prior art (such as CN112163536B-wavelet threshold function denoising method improved based on particle swarm algorithm), compared with patent CN112163536B, the embodiment of the present invention has the following differences and effects:

[0082] 1. The embodiment of the present invention optimizes three parameters, namely, the number of wavelet decomposition layers L, the threshold adjustment coefficient c, and the threshold function adjustment coefficient m. The final denoising effect is achieved under the coordinated optimization of the three parameters. Patent CN112163536B only optimizes the threshold, and does not consider setting the remaining parameters, or the remaining parameters still need to be set manually, which is very inconvenient and difficult to achieve the optimal denoising effect. The number of wavelet decomposition layers L determines the level of detail at which the signal is decomposed. The higher the number of decomposition layers, the narrower the signal band that can be separated, so as to better extract noise in different frequency bands. The threshold adjustment coefficient c determines the threshold, which determines which parts of the signal are considered to be noise, and thus are suppressed or filtered out. The threshold function adjustment coefficient m determines the threshold function, which determines the way the wavelet coefficients are processed during threshold processing, that is, how to process wavelet coefficients that exceed or are below the threshold.

[0083] 2. The optimization effect of the particle swarm optimization algorithm depends to a large extent on the fitness function, that is, only by accurately judging the optimization effect can we optimize better. In simple terms, only when we know what is good can we make something better. Patent CN112163536B uses the most commonly used SNR (Signal to Interference plus NoiseRatio) and RMSE (Root Mean Squared Error) as optimization indicators. However, in fact, these two parameters are difficult to obtain in practical applications, because the calculation formulas of these two parameters require the knowledge of the pure waveform without noise, which is unrealistic in actual situations. Generally, these two parameters can only appear in papers to evaluate and quantify the effect of denoising. Since the data in the papers are all simulation data, the original pure waveform can be known in advance, so it can be calculated. The evaluation index QNR of the embodiment of the present invention can be obtained without knowing the original pure signal, so it has strong applicability.

[0084] The method for adaptively extracting weak discharge signals of transformers under severe background noise interference of an embodiment of the present invention combines the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group as a particle, and adopts a particle swarm optimization algorithm to automatically optimize the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m, so that better parameters can be quickly obtained on site, with strong reliability and high practicality, especially for weak discharge signals under severe background noise interference. The original QNR is used as the fitness function of the particle swarm optimization algorithm, and there is no need to know the pure waveform of the original partial discharge signal, so it has strong applicability.

[0085] The effects of the embodiments of the present invention are verified by simulation and experiments as follows.

[0086] Simulation Verification

[0087] Studies have shown that the ultrasonic signal generated by partial discharge can be described by a double exponential decay oscillation model, which is expressed as follows:

[0088]

[0089] Among them, A m represents the amplitude, which is a random value between 5 and 10; τ represents the attenuation coefficient, represents the oscillation frequency, which is 100 microseconds here. The simulated partial discharge signal is as follows Figure 3 As shown, the total length is 20 milliseconds, and the f of each discharge pulse is c Is a random value between 60-150kHz.

[0090] By superimposing white noise with different signal-to-noise ratios on the pure partial discharge signal, the simulation of noisy partial discharge ultrasonic signals can be achieved. Figure 4 The PD signal with a signal-to-noise ratio of -5dB is shown.

[0091] The method of the embodiment of the present invention is used to Figure 4 Denoising is performed, and the denoising effect is as follows Figure 5 shown.

[0092] Experimental verification

[0093] During the annual maintenance of the standby phase converter transformer at a converter station, abnormal partial discharge was detected. The partial discharge detection was performed using an ultrasonic sensor placed on the outer wall of the front fuel tank. The voltage of the converter transformer was gradually increased during the detection process. The collected partial discharge ultrasonic signal is shown in Figure 1. Figure 6 shown.

[0094] The method of this embodiment is used to perform denoising, and the denoising result is as follows: Figure 7 The results of denoising using the existing method are shown in Figure 8 As shown. Figure 7 and Figure 8 It can be seen that the method of this embodiment can retain more waveform energy and has better waveform restoration performance.

[0095] The embodiment of the present invention also provides a transformer weak discharge signal adaptive extraction system under severe background noise interference, and the transformer weak discharge signal adaptive extraction system under severe background noise interference includes:

[0096] The particle swarm construction module is used to combine the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and use it as a particle. The positions of the three parameters in each particle are randomly assigned initial values ​​within the value range. At the same time, the initial speed of each particle is randomly assigned an initial value within the value range. N particles are randomly generated to establish a particle swarm.

[0097] The wavelet threshold denoising module is used to improve the threshold function and thresholding function in the wavelet threshold denoising method based on the parameters in the particles, use the improved threshold function and thresholding function to process the input signal A, traverse all particles, and obtain N results;

[0098] Particle update module, used to update the moving speed and position of each particle;

[0099] The fitness value calculation module is used to calculate the fitness value of each particle;

[0100] Individual historical optimal solution update module, used to update the individual historical optimal solution of each particle according to the fitness value;

[0101] The global optimal solution updating module is used to update the global optimal solution of the particle swarm according to the fitness value;

[0102] The output module is used to obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m as the optimal solution when the maximum number of iterations is reached or the global optimal solution of the particle no longer changes. The input signal A is processed accordingly and the denoising result is output.

[0103] An embodiment of the present invention further provides a computer device, including a processor and a storage medium, wherein a computer program is stored in the storage medium, and when the computer program is executed by the processor, the aforementioned method for adaptively extracting transformer weak discharge signals under severe background noise interference is implemented.

[0104] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed, the aforementioned method for adaptively extracting transformer weak discharge signals under severe background noise interference is implemented.

[0105] The above is only a specific implementation of the invention, but the protection scope of the invention is not limited thereto. Those skilled in the art should understand that the invention includes but is not limited to the contents described in the above specific implementation. Any modification that does not deviate from the functional and structural principles of the invention will be included in the scope of the claims.

Claims

1. The adaptive extraction method of transformer weak discharge signal under severe background noise interference is characterized by: The method for adaptively extracting transformer weak discharge signals under severe background noise interference includes: Step S1, combining the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and treating it as a particle, randomly assigning initial values ​​to the positions of the three parameters in each particle within a value range, and at the same time, randomly assigning initial values ​​to the initial speed of each particle within a value range, randomly generating N particles, and establishing a particle swarm; Step S2, based on the parameters in the particles, the threshold function and the thresholding function in the wavelet threshold denoising method are improved, the improved threshold function and the thresholding function are used to process the input signal A, and all particles are traversed to obtain N results; Step S3, updating the moving speed and position of each particle; Step S4, calculating the fitness value of each particle; Step S5, updating the individual historical optimal solution of each particle according to the fitness value; Step S6, updating the global optimal solution of the particle swarm according to the fitness value; Step S7, repeat the process from step S3 to step S6 until the maximum number of iterations is reached or the global optimal solution of the particle no longer changes, and obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m at this time as the optimal solution, and process the input signal A accordingly to output the denoising result.

2. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 1 is characterized in that: In step S2, the improved thresholding function is expressed as: Where, d s (j,L) is the function of the input signal A on the jth layer after L layers of wavelet decomposition, d' s (j,L) is the value of d s (j, L) is the output function after thresholding, T is the set threshold, and m is the custom thresholding function adjustment coefficient.

3. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 2 is characterized in that: In step S2, the improved threshold function is expressed as: T=c×max(d n (j,L)) Where c is the threshold adjustment coefficient, d n (j,L) represents the noise segment A in the input signal A N The function on the jth layer after L-layer wavelet decomposition, A N The first 10% or last 10% of signal A.

4. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 1 is characterized in that: In step S3, after t iterations, the three parameters L, c, and m of each particle are updated using the following formulas for the t+1th iteration: v L (t+1)=w·v L (t)+c1·r1·(p L_best (t)-L(t))+c2·r2·(g L_best (t)-L(t)) v c (t+1)=w·v c (t)+c1·r1·(p c_best (t)-c(t))+c2·r2·(g c_best (t)-c(t)) vm(t+1)=w·vm(t)+c1·r1·(pm _ best(t)-m(t))+c2·r2·(gm _ best(t)-m(t)) In the formula, v L 、v c 、v m are the speeds of L, c, and m respectively, w is the inertia weight, c1 and c2 are acceleration constants, r1 and r2 are random numbers between [0,1], and p L_best 、p c_best 、p m_best are the individual historical optimal solutions of L, c, and m, respectively, and g L_best ,g c_best ,g m_best are the global optimal solutions of L, c, and m respectively.

5. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 1 is characterized in that: In step S4, QNR is used as the fitness evaluation index of the particle, and the calculation formula of QNR is: Where l is the length of the signal, p is the number of delay cycles specified, is the delayed k-order autocorrelation coefficient of the noise residual. The smaller the QNR, the closer the noise residual is to white noise, indicating that the denoised signal is closer to the partial discharge signal.

6. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 1 is characterized in that: In step S5, if the fitness value of the current particle is better than the optimal value of the fitness in the particle's history, the individual historical optimal solution of the particle is updated; in step S6, if the fitness value of the current particle is better than the global optimal value of the current fitness, the global optimal solution of the entire particle swarm is updated.

7. The method for adaptively extracting transformer weak discharge signals under severe background noise interference according to claim 1 is characterized in that: In step S1, the value range of the wavelet decomposition layer number L is 4 to 20, the value range of the threshold adjustment coefficient c is 0.2 to 10, and the value range of the threshold function adjustment coefficient m is 2 to 2000.

8. The adaptive extraction system of transformer weak discharge signal under severe background noise interference is characterized by: The transformer weak discharge signal adaptive extraction system under severe background noise interference includes: The particle swarm construction module is used to combine the wavelet decomposition layer number L, the threshold adjustment coefficient c and the threshold function adjustment coefficient m into a parameter group and use it as a particle. The positions of the three parameters in each particle are randomly assigned initial values ​​within the value range. At the same time, the initial speed of each particle is randomly assigned an initial value within the value range. N particles are randomly generated to establish a particle swarm. The wavelet threshold denoising module is used to improve the threshold function and thresholding function in the wavelet threshold denoising method based on the parameters in the particles, use the improved threshold function and thresholding function to process the input signal A, traverse all particles, and obtain N results; Particle update module, used to update the moving speed and position of each particle; The fitness value calculation module is used to calculate the fitness value of each particle; Individual historical optimal solution update module, used to update the individual historical optimal solution of each particle according to the fitness value; The global optimal solution updating module is used to update the global optimal solution of the particle swarm according to the fitness value; The output module is used to obtain the wavelet decomposition layer number L, threshold adjustment coefficient c and threshold function adjustment coefficient m as the optimal solution when the maximum number of iterations is reached or the global optimal solution of the particle no longer changes. The input signal A is processed accordingly and the denoising result is output.

9. A computer device comprising a processor and a storage medium, wherein the storage medium stores a computer program, characterized in that: When the computer program is executed by a processor, the method for adaptively extracting transformer weak discharge signals under severe background noise interference as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method for adaptively extracting transformer weak discharge signals under severe background noise interference as claimed in any one of claims 1 to 7 is implemented.

Images