Switch cabinet partial discharge signal positioning method based on time difference positioning method

Abstract

The invention relates to a switch cabinet partial discharge signal positioning method based on a time difference positioning method, and belongs to the technical field of electrical engineering and signal processing. Aiming at the problem that the ultrasonic partial discharge signal has white noise, an improved wavelet threshold algorithm is adopted to construct a threshold function, and a continuous self-adaptive wavelet threshold denoising algorithm is adopted to carry out denoising processing on the partial discharge signal, so that the problems of excessive denoising and incomplete denoising existing in a traditional denoising algorithm are solved, effective separation of the signal and the noise is realized, and the denoising efficiency is improved. And the performance of the denoising algorithm is improved. In order to obtain a time delay value between each path of signals, a generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting is used for substituting a time delay estimation value into a switch cabinet partial discharge positioning equation set, and an optimal equation is obtained. According to the partial discharge optimal solution algorithm based on DE-PSO, the optimization equation is solved, the spatial coordinate position generating the partial discharge signal is positioned, the defects of parameter setting and threshold selection of the optimization algorithm are overcome, and the positioning precision of the partial discharge signal is improved.

Description

Technical Field

[0001] The present invention belongs to the technical fields of electrical engineering and signal processing, and relates to a method for locating partial discharge signals in switchgear based on time difference positioning method. Background Art

[0002] At present, the detection of partial discharge (PD) in switchgear generally relies on traditional regular detection mechanisms. However, this mechanism has obvious limitations. Its maintenance cycle is relatively long, and potential faults are likely to occur during the interval between two standard maintenance periods, and it fails to detect and handle them in time, which may lead to more serious equipment damage and even system power outage accidents. At the same time, unnecessary over-repair of equipment that is still in a healthy state not only wastes valuable maintenance resources but also increases operating costs. With the continuous development and expansion of the power system, the number of switchgears has increased sharply, and their internal structures have become increasingly complex, which undoubtedly brings unprecedented challenges to the research and detection of partial discharge.

[0003] The complex electrical and mechanical structures inside switchgear often lead to serious noise interference in the collected partial discharge signals. These noises may come from the external environment, mechanical vibrations inside the equipment, and electromagnetic interference, etc. They will greatly affect the type recognition and feature extraction of partial discharge signals. When traditional denoising algorithms are used to process partial discharge signals, they often face the dilemma of over-denoising and incomplete denoising. On the one hand, over-denoising may cause useful information in the signal to be deleted by mistake, affecting subsequent signal analysis and fault diagnosis; on the other hand, incomplete denoising cannot effectively eliminate the interference of noise, resulting in an unsatisfactory separation effect between the signal and the noise, and the overall performance of the denoising algorithm still needs to be further improved. Patent CN116626457A discloses a method and system for locating ultra-high frequency partial discharge of transformers based on SSA optimization. This patent uses the SSA algorithm to solve for multiple positioning solutions, uses the K-means algorithm to perform clustering analysis on the obtained positioning solutions to obtain the clustering center, and uses the SSA algorithm to calculate the optimal solution with the smallest sum of distances to the clustering center to obtain the positioning point of the ultra-high frequency partial discharge source of the transformer. However, this patent still has problems with the setting of optimization algorithm parameters and threshold selection, and cannot further improve the positioning accuracy and efficiency. Patent CN110927543A discloses a method for estimating the time difference of ultrasonic signals of partial discharge in power equipment. This patent decomposes the original signal into multiple intrinsic mode functions IMF with different frequency components through the CEEMDAN algorithm, calculates the autocorrelation of each IMF through the correlation coefficient method to identify the demarcation point between the noise component and the partial discharge ultrasonic signal, sets the noise threshold according to the calculated demarcation point, and performs wavelet threshold denoising on the IMF less than the noise threshold to obtain the denoised partial discharge ultrasonic signal. However, this patent still has problems with the accuracy of time difference estimation and cannot further improve the accuracy of time difference estimation.

[0004] Therefore, the prior art has the following disadvantages: In the denoising of partial discharge signals, traditional denoising algorithms have problems of over-denoising and incomplete denoising, and cannot effectively separate signals from noise. The performance of denoising algorithms needs to be improved. Existing time-delay estimation algorithms are affected by complex on-site environments and various interferences, resulting in poor positioning accuracy or even positioning failure. There is an urgent need for a new time-delay estimation algorithm to improve positioning accuracy. Existing algorithms have defects in optimizing algorithm parameter settings and threshold selection, and cannot further improve positioning accuracy and efficiency. Existing algorithms have deficiencies in the accuracy of time difference estimation and cannot further improve the accuracy of time difference estimation.

[0005] Aiming at the problems of over-denoising and incomplete denoising in the denoising of partial discharge signals, the present invention proposes a method for denoising partial discharge signals of switchgear based on an improved wavelet threshold algorithm. This method first collects the partial discharge signals inside the switchgear through an ultrasonic partial discharge signal acquisition circuit, and then preprocesses the signals using a filter amplification circuit to improve the signal-to-noise ratio of the signals. Next, an improved wavelet threshold algorithm is used to denoise the preprocessed signals. By optimizing parameters such as the selection of wavelet basis functions, the setting of thresholds, and the shape of threshold functions, this algorithm realizes effective denoising of partial discharge signals while avoiding problems of over-denoising and incomplete denoising.

[0006] In addition, aiming at the problem of difficult localization of partial discharge signals, the present invention combines the advantages of phase transform (PHAT) weighting and smoothed coherence transform (SCOT) weighting, aiming to improve the anti-noise ability and accuracy of the generalized cross-correlation time-delay estimation algorithm. PHAT weighting enhances the correlation of signals through phase transform, while SCOT weighting reduces the influence of noise through smoothed coherence transform. Combining the two can further improve the performance of time-delay estimation, especially in a low signal-to-noise ratio environment. Therefore, the generalized cross-correlation time-delay estimation algorithm based on PHAT-SCOT joint weighting proposed by the present invention combines the advantages of PHAT weighting and SCOT weighting, and can more accurately estimate the time-delay value of partial discharge signals between different sensors. Then, substituting these time-delay estimation values into the established partial discharge localization equation set of the switchgear cabinet, the optimal spatial coordinate position is obtained by solving this equation set. To solve this optimization equation, the present invention further designs a partial discharge optimal solution algorithm based on differential evolution algorithm (DE) and particle swarm optimization algorithm (PSO). This algorithm realizes the rapid solution of the optimization equation by combining the global search ability of the DE algorithm and the local convergence speed of the PSO algorithm, and finally accurately locates the spatial coordinate position where the partial discharge signal is generated. Summary of the Invention

[0007] In view of the problems of poor denoising effect, over-denoising or incomplete denoising of partial discharge signals, as well as low localization accuracy of partial discharge signals and inaccurate time-delay estimation in the prior art. The purpose of the present invention is to provide a method for denoising and localizing partial discharge signals of switchgear cabinets by combining an improved wavelet threshold algorithm and a time-difference localization method.

[0008] To achieve the above purpose, the present invention first uses an ultrasonic partial discharge signal acquisition circuit to convert ultrasonic waves into electrical signals and filter and amplify the electrical signals. Then, aiming at the problem of white noise existing in ultrasonic partial discharge signals, the traditional wavelet threshold denoising algorithm is improved, a new threshold function is constructed, and a continuous adaptive wavelet threshold denoising algorithm is proposed to denoise the partial discharge signals. To obtain the time-delay values between various signals, a generalized cross-correlation time-delay estimation algorithm based on PHAT-SCOT joint weighting is designed, the time-delay estimation values are substituted into the established partial discharge localization equation set of the switchgear cabinet to obtain an optimization equation, and a partial discharge optimal solution algorithm based on DE-PSO is designed to solve the optimization equation, and finally the spatial coordinate position where the partial discharge signal is generated is located. The present invention specifically provides the following technical solutions:

[0009] A method for locating partial discharge signals in switchgear based on time difference positioning method, comprising the following steps:

[0010] Step 1: Use an ultrasonic partial discharge signal acquisition circuit to collect signals and filter and amplify the signals;

[0011] Step 2: Determine the wavelet basis function by comparing the effects after denoising with different wavelet basis functions and different decomposition levels;

[0012] Step 3: Calculate the adjustment parameter k of each wavelet decomposition layer and determine the threshold function corresponding to each layer of wavelet coefficients;

[0013] Step 4: For a given threshold λ 0 , calculate the unbiased likelihood estimate value, and then minimize this threshold to obtain the optimal threshold of each wavelet decomposition layer;

[0014] Step 5: Substitute the adjustment parameter k into the threshold function to determine the threshold function of each layer, and process the wavelet coefficients to obtain the wavelet coefficient estimate value;

[0015] Step 6: Perform wavelet inverse reconstruction on the processed wavelet coefficients to obtain the denoised partial discharge signal;

[0016] Step 7: Establish a partial discharge positioning equation set for the switchgear;

[0017] Step 8: Use a generalized cross-correlation time-delay estimation algorithm based on joint weighting of phase transform and smoothed coherence transform (PHAT-SCOT) to estimate the time delay;

[0018] Step 9: Substitute the time delay estimate value into the established partial discharge positioning equation set for the switchgear to obtain an optimization equation;

[0019] Step 10: Use a partial discharge optimal solution algorithm based on differential evolution algorithm and particle swarm optimization algorithm (Differential Evolution Algorithm - Particle Swarm Optimization, DE-PSO) to solve the optimization equation to obtain the spatial coordinate position of the partial discharge signal.

[0020] Further, the ultrasonic partial discharge signal acquisition circuit in Step 1 includes: an autotransformer, a step-up transformer, a protection resistor, a coupling capacitor, a measuring impedance, a switchgear, a discharge model, an ultrasonic sensor, a preamplifier, and an oscilloscope, and the preamplifier includes a filter and a signal amplifier;

[0021] Step 1 includes:

[0022] Step 102: Filter the acquired signal through the filter to remove high-frequency noise;

[0023] Step 103: Amplify the filtered signal through the signal amplifier to increase the signal strength.

[0024] Furthermore, the threshold function in Step 3 is a continuous adaptive wavelet threshold function, which adaptively adjusts the characteristics of the threshold function according to the characteristics of the partial discharge signal at different wavelet decomposition levels to adapt to the partial discharge signal denoising under different noise conditions. The continuous adaptive wavelet threshold function is expressed as the following formula:

[0025]

[0026] In the formula, w x,k is the wavelet coefficient after being processed by the threshold function, x is the wavelet coefficient, λ is the threshold, and k is the adjustment parameter of the threshold function.

[0027] Furthermore, the value of the adjustment parameter k in the continuous adaptive wavelet threshold function is used to change the variation trend of the threshold function.

[0028] Furthermore, for different partial discharge signal decompositions, the threshold function of each signal decomposition layer is determined by adjusting the parameter k to adapt to the characteristics of each partial discharge signal and the proportion of noise energy in each decomposition layer.

[0029] Furthermore, the adjustment parameter k is determined according to the ratio of the signal to the noise, and its calculation formula is:

[0030]

[0031] In the formula, E di is the energy of the noisy signal in the i-th layer of wavelet decomposition, E ni is the energy of the noise in the i-th layer of wavelet decomposition, E di and E ni are calculated based on the scale coefficient and the detail coefficient;

[0032] Assume that when all the signals in the i-th layer are noise, E ni ≈E di , and at this time k is 1; when the signals in the i-th layer are completely free of noise, E ni = 0, k is 0; 0 ≤ k ≤ 1.

[0033] Furthermore, Step 7 is specifically as follows:

[0034] Assume that S(x, y, z) is the coordinate of the partial discharge source inside the switchgear, and four ultrasonic sensors are placed for partial discharge detection. The positions of the sensors are P 1 (x 1 , y1 , z 1 ), P 2 (x 2 , y 2 , z 2 ), P 3 (x 3 , y 3 , z 3 ), P 4 (x 4 , y 4 , z 4 ), construct a ternary quadratic equation system to solve for the position of the discharge source S;

[0035] Assume that the sound speed v represents the ultrasonic wave propagation speed, and the arrival time of the ultrasonic signal is t 1 , t 2 , t 3 , t 4 , and use the method of constructing an optimization problem for indirect solution to obtain the objective function L expressed as the following formula:

[0036]

[0037] where L ≥ 0, and x, y, z are the solutions of the equation system if and only if L = 0;

[0038] The constructed optimization problem is expressed as the following formula:

[0039]

[0040] Use the optimal solution algorithm to solve the above optimization problem to find the optimal solution of x, y, z values that meet the requirements to obtain the partial discharge source coordinates S(x, y, z) and realize partial discharge source positioning.

[0041] Furthermore, step 8 includes the following steps:

[0042] Step 801, perform cross-correlation operation on signal x 1 (t) and signal x 2 (t) to obtain R 12 (τ);

[0043] Step 802, perform generalized weighting on the cross-correlation result R 12 (τ) using the PHAT-SCOT joint weighting factor to obtain the power spectral density function G 12 (ω);

[0044] Step 803, perform inverse Fourier transform on the power spectral density function G 12 (ω) and perform peak detection to obtain the time delay estimation value.

[0045] Further, step 9 is specifically as follows:

[0046] Let the signals x 1 (t) and x 2 (t) be represented by the following mathematical models:

[0047]

[0048] In the formula: x 1 (t) and x 2 (t) are time-delay estimation signals; s(t) is the signal source; m 1 (t) and m 2 (t) are white noises superimposed in the time-delay estimation signals; T is the time delay between the above time-delay estimation signals;

[0049] The cross-correlation function R 12 (τ) of the above two time-delay estimation signals is expressed as:

[0050] R 12 (τ) = E[x 1 (t)x 2 (t - τ)]

[0051] The cross-correlation function R 1 (t) and x 2 (t) and its cross-power spectral density function G 12 (ω) are related by the following expression: 12 (ω) is expressed as:

[0052]

[0053] The generalized cross-correlation function is expressed as:

[0054]

[0055] In the formula: ψ(ω) is the weighting function;

[0056] The expression of the PHAT weighting factor is as follows:

[0057]

[0058] In the formula: is the cross-power spectrum of the signals x 1 (t) and x 2 (t);

[0059] The expression of the SCOT weighting factor is as follows:

[0060]

[0061] In the formula: is the cross-power spectrum of the signals x 1(t) self-power spectrum; is the signal x 2 (t) self-power spectrum;

[0062] The PHAT-SCOT combined weighting factor expression is as follows:

[0063]

[0064] In the formula: is the signal x 1 (t), x 2 (t) cross-power spectrum of the received noise; ρ is the correlation factor.

[0065] Furthermore, step 10 includes the following steps:

[0066] Step 1001, initialize two parallel iterative subpopulations of DE and PSO, and find the global optimal solution best through the DE algorithm DE , and find the global optimal solution best through the PSO algorithm PSO ;

[0067] Step 1002, compare best DE and best PSO the magnitudes of the two optimal solutions to update the velocity formula of the PSO algorithm, and stop iterating after solving to the maximum number of iterations in sequence;

[0068] Step 1003, compare the global optimal solutions of the DE algorithm and the PSO algorithm generated in each iteration, and take the smaller of the two as the global optimal solution of the DE-PSO algorithm.

[0069] The beneficial effects of the present invention are as follows:

[0070] Compared with the prior art, the present invention provides a method for denoising and positioning partial discharge signals of switchgear cabinets combining an improved wavelet threshold algorithm and a time difference positioning method, having the following beneficial effects: By using the improved wavelet threshold algorithm, the problems of excessive denoising and incomplete denoising existing in traditional denoising algorithms can be effectively overcome, the effective separation of signals and noise is realized, and the performance of the denoising algorithm is improved; A generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT combined weighting is designed, which can effectively overcome the influence of complex on-site environments and various interferences, improve the accuracy of time delay estimation, and thus improve the positioning accuracy of partial discharge signals; A continuous adaptive wavelet threshold function is constructed, and by adjusting the adjustment parameter k of each wavelet decomposition layer, the characteristics of the threshold function are adaptively adjusted, which can better adapt to the denoising of partial discharge signals under different noise conditions and improve the denoising effect; A local discharge optimal solution algorithm based on DE-PSO is designed, which can effectively solve the defects existing in the optimization algorithm parameter setting and threshold selection of existing algorithms, and improve the positioning accuracy and efficiency.

[0071] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be learned from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail preferably with reference to the accompanying drawings, where:

[0073] Figure 1 is the circuit diagram for collecting partial discharge signals in an embodiment of the present invention;

[0074] Figure 2 is the flow chart of the improved wavelet threshold algorithm in an embodiment of the present invention;

[0075] Figure 3 is the flow chart of the generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting in an embodiment of the present invention;

[0076] Figure 4 is the flow chart of the DE-PSO algorithm in an embodiment of the present invention;

[0077] Figure 5 is the flow chart of an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0078] The following illustrates the embodiments of the present invention through specific specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the drawings provided in the following embodiments only illustrate the basic concept of the present invention schematically. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0079] Among them, the drawings are only for illustrative purposes, showing only schematic diagrams, not physical diagrams, and should not be construed as limiting the present invention; for better illustrating the embodiments of the present invention, some components in the drawings will be omitted, enlarged, or reduced, and do not represent the dimensions of actual products; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0080] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the accompanying drawings. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the accompanying drawings are only for illustrative purposes and cannot be construed as a limitation on the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0081] 1. Switchgear partial discharge signal acquisition

[0082] Please refer to Figure 1 , which is the circuit diagram for partial discharge signal acquisition in the embodiments of the present invention;

[0083] The main structure of the acquisition circuit includes an autotransformer, a step-up transformer, a protection resistor, a coupling capacitor, a measuring impedance, a switchgear, a discharge model, an ultrasonic sensor, a preamplifier, and an oscilloscope. The preamplifier includes a filter and a signal amplifier.

[0084] Currently, the most commonly used sensor in the partial discharge ultrasonic detection method is the piezoelectric ultrasonic sensor. The working principle of the piezoelectric ultrasonic sensor is that the piezoelectric element in it senses the pressure wave and converts the vibration signal into an electrical signal. Usually, the sensor is placed at the outer wall gap of the device. After the sensor collects the ultrasonic partial discharge signal, it is preprocessed by a filter and a signal amplifier. The collected signal is filtered by the filter to remove high-frequency noise; the filtered signal is amplified by the signal amplifier to increase the signal strength. In order to reduce the interface loss of the sound wave during the signal acquisition process, a layer of coupling agent is usually applied between the ultrasonic sensor and the outer wall of the device to enhance the acoustic coupling.

[0085] Since the partial discharge signal is a transient pulse signal and the discharge time is very short, a high-frequency oscilloscope with a bandwidth of 200 MHz and a maximum sampling rate of 1 GS / s is selected in this paper to collect the ultrasonic partial discharge signal. During signal acquisition, the ultrasonic sensor module is closely attached to the outer wall gap of the switchgear, and the partial discharge signal generated by the simulated discharge inside the cabinet can be detected. Then, it is input into the oscilloscope through the signal preprocessing module, and finally the data is imported into the computer. Each experiment records the data once as the basic data for subsequent pattern recognition.

[0086] Please refer to Figure 2 , which is the flow chart of the improved wavelet threshold algorithm in the embodiments of the present invention;

[0087] 2. Improved wavelet threshold denoising algorithm

[0088] 2.1 Threshold function construction

[0089] The characteristics of the wavelet threshold function are very important for ensuring the denoising effect of the algorithm. When the threshold function is continuous in the wavelet domain, oscillations and blurring of the signal after denoising can be avoided. When the threshold function has y = x as its asymptote, constant deviation can be avoided, and the original signal information after denoising can be better retained. In order to better remove the white noise in the partial discharge signal and retain the detailed characteristics of the partial discharge signal as much as possible, the present invention constructs a continuous adaptive wavelet threshold function, which adaptively adjusts the characteristics of the threshold function according to the characteristics of the partial discharge signal at different wavelet decomposition levels, and can better adapt to the denoising of partial discharge signals under different noise conditions. The formula of the continuous adaptive wavelet threshold function is as follows:

[0090]

[0091] where w x,k is the wavelet coefficient processed by the threshold function, x is the wavelet coefficient, λ is the threshold, and k is the adjustment parameter of the threshold function. Next, the continuity, asymptotics, and constant difference of the function are proved and analyzed.

[0092] (1) The proof of the continuity of the threshold function is as follows:

[0093] At when

[0094]

[0095] When x → -λ - when

[0096]

[0097] (2) Analysis of the asymptotics of the threshold function:

[0098] When x > λ

[0099]

[0100] When x < λ

[0101]

[0102] (3) Analysis of the constant difference of the threshold function

[0103] When x → +∞

[0104]

[0105] When x → -∞

[0106]

[0107] According to the above formula, it can be seen that the proposed threshold function is continuous at both positive and negative thresholds, and the asymptote slope is equal to 1 when x→+∞. At the same time, by adjusting the value of k, the proportion of noise eliminated by the function can be changed to eliminate the constant deviation. In summary, the designed continuous adaptive threshold function meets the requirements of the threshold function.

[0108] 2.2 Threshold Function Characteristics and Selection of Adjustment Parameters

[0109] There is an adjustment parameter k in the formula of the continuous adaptive threshold function proposed by the present invention. The value of k will change the change trend of the threshold function. By adjusting the adjustment parameter k, the shrinkage characteristics of the threshold function can be changed. Since the threshold function is continuous near the threshold, near the threshold, it is not easy to distinguish the wavelet coefficients of the white noise part and the useful signal part in the partial discharge signal. This part can be adjusted by changing the value of k to adjust the change trend of the threshold function, and then change the shrinkage degree of the threshold function to change the proportion of noise removed by the continuous adaptive wavelet threshold denoising algorithm. According to the characteristics of the continuous adaptive threshold function, the smaller k is, the more suitable it is for partial discharge signals with a small proportion of noise removed; the larger k is, the more suitable it is for partial discharge signals with a large proportion of noise removed. The continuous adaptive wavelet threshold denoising algorithm in this paper is proposed for partial discharge signals. Due to the high-frequency characteristics of partial discharge signals, it is easy to be confused with white noise. Therefore, it is necessary to adopt threshold functions with different adjustment parameters in each wavelet decomposition layer for adaptive denoising.

[0110] For different decompositions of the partial discharge signal, the adjustment parameter k in the continuous adaptive threshold function can be adjusted according to its characteristics and the proportion of noise energy in different layers to determine the threshold function most suitable for different decomposition layers of the signal, so as to achieve the best denoising effect. For the decomposition layer with a relatively large proportion of noise energy, a threshold function with a larger k value should be selected, which has a better effect on removing strong noise. For the decomposition layer with a relatively small proportion of noise energy, a threshold function with a smaller k value should be selected, which pays more attention to retaining the useful part in the partial discharge signal. As the proportion of the useful signal energy in the partial discharge signal increases, the selected value of k gradually decreases. Therefore, the value of k can be determined according to the ratio of the signal to the noise. The calculation formula for the value of k is:

[0111]

[0112] where E di is the energy of the noisy signal in the i-th layer of wavelet decomposition, and E ni is the energy of the noise in the i-th layer of wavelet decomposition. E di and E ni can be calculated according to the scale coefficient and the detail coefficient. Assuming that when all the signals in the i-th layer are noise, E ni ≈E di , and at this time, k reaches the maximum value of 1. When the signal in the i-th layer is completely free of noise, E ni= 0, k takes the minimum value of 0. Under normal operating conditions, the algorithm can adaptively obtain the value of k, and the value range of k is 0 ≤ k ≤ 1.

[0113] For a given threshold λ 0 , the unbiased likelihood estimate value is calculated, and then the threshold is minimized to obtain the optimal threshold for each wavelet decomposition layer;

[0114] Substitute the adjustment parameter k into the threshold function to determine the threshold function for each layer, and process the wavelet coefficients to obtain the estimated wavelet coefficient values;

[0115] Perform inverse wavelet reconstruction on the processed wavelet coefficients to obtain the denoised partial discharge signal.

[0116] 3. Time difference location method

[0117] 3.1 Establish the partial discharge location equation set for switchgear

[0118] During the operation of high-voltage power equipment such as switchgear, partial discharge will generate ultrasonic signals. By deeply analyzing the time difference of ultrasonic signals arriving at different sensors, the accurate identification of the partial discharge location inside the switchgear can be achieved.

[0119] Assume that S(x, y, z) is the coordinate of the partial discharge source inside the switchgear. Four ultrasonic sensors are placed for partial discharge detection, and the sensor positions are P 1 (x 1 , y 1 , z 1 ), P 2 (x 2 , y 2 , z 2 ), P 3 (x 3 , y 3 , z 3 ), P 4 (x 4 , y 4 , z 4 ). A ternary quadratic equation set is constructed to solve the position of the partial discharge source S.

[0120] Assume that the sound speed v represents the ultrasonic propagation speed. In this paper, v = 340 m / s, and the arrival times of the ultrasonic signals are t 1 , t 2 , t 3 , t 4 . Then the following equation set can be listed:

[0121]

[0122] After taking the square, the position information of the discharge source can be obtained directly by solving the above equations. However, considering that there may be imaginary solutions or even no solutions in the process of solving, an indirect solution method is adopted by constructing an optimization problem.

[0123] Equation (9) can be transformed into:

[0124]

[0125] In the formula:

[0126] Δt1—the time difference between t2 and t1;

[0127] Δt2—the time difference between t3 and t1;

[0128] Δt3—the time difference between t4 and t1;

[0129] L1—the value obtained by subtracting the distance difference calculated by the time difference method from the actual distance difference between the P2, P1 sensors and the discharge source S;

[0130] L2—the value obtained by subtracting the distance difference calculated by the time difference method from the actual distance difference between the P3, P1 sensors and the discharge source S;

[0131] L3—the value obtained by subtracting the distance difference calculated by the time difference method from the actual distance difference between the P4, P1 sensors and the discharge source S.

[0132] Thus, the objective function L is expressed as the following formula:

[0133]

[0134] where L≥0, and x, y, z are the solutions of the equations if and only if L = 0.

[0135] When applied to actual power equipment, there is a definite range for the possible occurrence area of partial discharge. Based on this, an optimization problem can be constructed as follows:

[0136]

[0137] Use the optimal solution algorithm to solve the above optimization problem and find the optimal solution of the x, y, z values that meet the requirements, then the coordinates of the partial discharge source S(x, y, z) can be obtained, thus realizing the localization of the partial discharge source.

[0138] Please refer to Figure 3 , which is the flow chart of the generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting in the embodiment of the present invention;

[0139] 3.2 Generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting

[0140] The specific steps of the generalized cross - correlation time - delay estimation algorithm based on PHAT - SCOT joint weighting are as follows:

[0141] (1) Perform cross - correlation operation on signal x 1 (t) and signal x 2 (t) to obtain R 12 (τ);

[0142] (2) Generalize the weighting of the cross - correlation result R 12 (τ) using the PHAT - SCOT joint weighting factor to obtain the power spectral density function G 12 (ω);

[0143] (3) Perform the inverse Fourier transform on the power spectral density function G 12 (ω) and perform peak detection to obtain the time - delay estimation value.

[0144] Assume that the mathematical models of signals x 1 (t), x 2 (t) are expressed as follows:

[0145]

[0146] In the formula:

[0147] x 1 (t), x 2 (t) - Time - delay estimation signals;

[0148] s(t) - Signal source;

[0149] m 1 (t), m 2 (t) - White noise superimposed on the signal;

[0150] T - Time delay between the two signals.

[0151] The cross - correlation function R 12 (τ) of the two - path signals can be expressed as:

[0152] R 12 (τ) = E[x 1 (t)x 2 (t - τ)] (14)

[0153] It can be known from the Wiener - Khintchine theorem that the relationship expression between the cross - correlation function R 1 (τ) of signals x 2 (t), x 12 (τ) and its cross - power spectral density function G 12 (ω) is:

[0154]

[0155] Then the final generalized cross-correlation function can be expressed as:

[0156]

[0157] Where:

[0158] ψ(ω) - - Weighting function.

[0159] The phase transform (PHAT) weighting factor is equivalent to a whitening filter, which can effectively suppress noise. However, when the signal energy is small, the denominator tends to zero, resulting in an increase in error. The expression of this weighting factor is as follows:

[0160]

[0161] Where:

[0162] - - Signal x 1 (t), x 2 (t) cross-power spectrum.

[0163] The smoothed coherence transform (SCOT) weighting factor can consider the influence of both signals at the same time and can effectively solve the influence of signal fluctuations on time-delay estimation in the case of low signal-to-noise ratio. However, when the power spectral densities of the two signals are equal, it will broaden the peak of the correlation function, resulting in incorrect estimation. The expression of this weighting factor is as follows:

[0164]

[0165] Where:

[0166] - - Signal x 1 (t) auto-power spectrum;

[0167] - - Signal x 2 (t) auto-power spectrum.

[0168] This paper adopts the PHAT-SCOT joint weighting factor, which can strengthen the sharpening effect of the cross-correlation function and obtain a more accurate time-delay estimation. The expression of this weighting factor is as follows:

[0169]

[0170] Where:

[0171] - - Signal x 1 (t), x 2 (t) cross-power spectrum of the received noise;

[0172] ρ —— correlation factor.

[0173] The value of ρ is adjusted accordingly with the change of the environmental signal-to-noise ratio, and the specific value needs to be optimized through multiple experimental tests.

[0174] Please refer to Figure 4 , which is the flow chart of the DE-PSO algorithm in the embodiment of the present invention;

[0175] 3.3 Optimal solution algorithm for partial discharge based on DE-PSO

[0176] The core operations of the differential evolution algorithm (DE) include two steps: mutation and crossover. The algorithm has strong robustness and outstanding performance in terms of search accuracy, especially good at local search. The particle swarm optimization algorithm (PSO) is a global search algorithm. Based on the concept of swarm intelligence, it can quickly achieve global convergence. However, it is prone to falling into local optimum in the later stage of iteration, and the particle diversity decreases in the later stage, resulting in premature convergence. The present invention uses the DE-PSO algorithm to solve the partial discharge optimization equation, which can not only give full play to the strong ability of the DE algorithm in local search, but also utilize the advantage of the fast global convergence speed of the PSO algorithm, and can effectively improve the solution efficiency and accuracy.

[0177] The DE algorithm first initializes the population, and screens out individuals with higher fitness through the mutation, crossover and selection mechanisms among individuals, so as to generate the next generation of population, and continuously repeats this process until a solution that meets the conditions is found. The specific steps are as follows:

[0178] (1) Initialize the population

[0179] Initialize the basic parameters, including the number of iterations G, the upper limit of the search space and the lower limit space dimension N, population size N P , mutation factor F and crossover factor C R . The initial population is randomly generated in , where i = 1, 2,..., N P , j = 1, 2,..., N, and the expression is as follows:

[0180]

[0181] In the formula:

[0182] —— The j-th dimensional vector of the i-th individual in the 0-th generation;

[0183] rand(0, 1) —— Random number uniformly distributed in the interval [0, 1].

[0184] (2) Mutation operation

[0185] For each individual x in the population by the differential method i mutate to produce the corresponding mutant offspring v i , and there are the following five ways:

[0186] ν i = x r1 + F × (x r2 - x r3 ) (21)

[0187] ν i = x best + F × x r1 - x r2 ) (22)

[0188] DE / rand-to-best / l: ν i = x i + F × (x best - x i ) + F × (x r1 - x r2 ) (23)

[0189] DE / rand / 2: ν i = x r1 + F × (x r2 - x r3 ) + F × (x r4 - x r5 ) (24)

[0190] DE / best / 2: ν i = x best + F × (x r1 - x r2 ) + F × (x r3 - x r4 ) (25)

[0191] In the formula:

[0192] r 1 , r 2 , r 3 , r 4 , r 5 —— random numbers in the interval [1, N p , and r 1 ≠ r 2 ≠ r 3 ≠ r 4 ≠ r 5 ≠ i.

[0193] (3) Crossover operation

[0194] To maintain the diversity of the population, the mutant offspring ν iand each individual x i perform crossover to generate corresponding crossover offspring u i , and the calculation formula is as follows:

[0195]

[0196] In the formula:

[0197] r j —— a random number in the interval [0, 1];

[0198] j rand —— a random integer, taking 1, 2, …, N.

[0199] (4) Selection operation

[0200] Adopt one-to-one greedy selection, that is, when the offspring solution vector u i is better than the parent solution vector x i , then replace the parent solution vector with the offspring solution vector in the next generation, otherwise still preserve the parent solution vector. The expression is as follows:

[0201]

[0202] where f(u i ) is the fitness of individual u i , and f(x i ) is the fitness of individual x i .

[0203] The PSO algorithm first randomly scatters the initial particles in the solution space. These particles, under the control of weights, continuously update and adjust their positions and velocities according to their own local and global optimal values, so as to find the optimal solution. The specific steps are as follows:

[0204] (1) Initialize the particle swarm

[0205] Initialize the basic parameters, including the space dimension D, population size P, number of iterations T, inertia coefficient w, learning factors c1 and c2, maximum velocity ν max and minimum velocity ν min , maximum position p max and minimum position p min . The initial population is randomly generated in , where i = 1, 2, …, P, j = 1, 2, …, D, and the expression is as follows:

[0206] p i = p min + rand(0, 1)(p max - p min ) (28)

[0207] νi = ν min + rand(0,1)(ν max - ν min ) (29)

[0208] Wherein:

[0209] p i —— The position of the i-th particle;

[0210] ν i —— The velocity of the i-th particle.

[0211] (2) Velocity and position update

[0212] Assume that the position and velocity of the i-th particle at the t-th iteration are p i (t), ν i (t). Up to the t-th generation, the local optimal position searched by particle i is denoted as p best (t), and the calculation formula is as follows:

[0213]

[0214] Wherein:

[0215] —— The position of the i-th D-dimensional particle at the t-th iteration.

[0216] The global optimal position searched by all particles in the population is denoted as G best (t), and the calculation formula is as follows:

[0217]

[0218] When the algorithm iterates to the (t + 1)-th time, the position and velocity update formulas of the i-th particle are as follows:

[0219] p i (t + 1) = p i (t) + ν i (t + 1) (32)

[0220] ν i (t + 1) = wν i (t) + c 1 r 1 (p best (t) - p i (t)) + c 2 r 2 (G best (t) - p i (t)) (33)

[0221] (3) Iteration termination condition

[0222] If the current result does not meet the termination condition, return to step (2) and recalculate the fitness of the particles. If the termination condition is met, output the global optimal value of the particles, and the expression is as follows:

[0223]

[0224] In the formula:

[0225] f(p i (t + 1)) —— The fitness of individual p i (t + 1);

[0226] f(p i (t)) —— The fitness of individual p i (t).

[0227] Based on the above analysis of the DE and PSO algorithms, this paper designs a DE - PSO algorithm by combining the advantages of the two algorithms. The algorithm steps are as follows:

[0228] (1) Initialize two parallel iterative sub - populations of DE and PSO, and find the global optimal solution best DE through the DE algorithm, and find the global optimal solution best PSO through the PSO algorithm;

[0229] (2) Compare the magnitudes of the two optimal solutions best DE and best PSO to update the velocity formula of the PSO algorithm, and stop the iteration after solving up to the maximum number of iterations;

[0230] (3) Compare the global optimal solutions of the DE algorithm and the PSO algorithm generated in each iteration, and take the smaller of the two as the global optimal solution of the DE - PSO algorithm.

[0231] In terms of the mutation operation, introduce a linear simulated annealing strategy, set λ as the linear annealing factor, and the calculation formula is as follows:

[0232]

[0233] In the formula:

[0234] t —— The current iteration number;

[0235] T max —— The maximum number of iterations.

[0236] Combining the characteristics of strong global search ability of DE / rand / 1 and strong local search ability of DE / best / 1, the mutation operation calculation formula is as follows:

[0237] ν i = λxr1 +(1 - λ)x best +F×(x r2 -x r3 ) (36)

[0238] where λ ∈ [0, 1]. When λ = 0, the above formula is the DE / rand / 1 mutation method; when λ = 1, the above formula is the DE / best / 1 mutation method.

[0239] In terms of velocity update, for each iteration, find the optimal fitness value best of the DE sub - population DE and the optimal fitness value best of the PSO sub - population PSO , and judge the two optimal fitness values. If the optimal fitness value of the DE sub - population is less than that of the PSO sub - population, then replace G best (t) in the velocity update formula (24) with the global optimal position G bestDE (t) of the DE population, aiming to guide the particles to approach the optimal individual in the global population. The velocity update formula is as follows:

[0240] ν i (t + 1) = wν i (t)+c 1 r 1 (p best (t)-p i (t))+c 2 r 2 (G bestDE (t)-p i (t)), best DE ≤ best PSO (37)

[0241] ν i (t + 1) = wν i (t)+c 1 r 1 (p best (t)-p i (t))+c 2 r 2 (G bestPSO (t)-p i (t)), best DE > best PSO (38)

[0242] Please refer to Figure 5, which is the flowchart of the embodiment of the present invention; First, the ultrasonic partial discharge signal acquisition circuit is used to convert ultrasonic waves into electrical signals and filter and amplify the electrical signals. Then, aiming at the problem of white noise existing in the ultrasonic partial discharge signal, the traditional wavelet threshold denoising algorithm is improved, a new threshold function is constructed, and the continuous adaptive wavelet threshold denoising algorithm is proposed to denoise the partial discharge signal. To obtain the time delay values between various signals, the generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting is designed, and the time delay estimation values are substituted into the established local discharge positioning equations of the switchgear cabinet to obtain the optimization equation. The local discharge optimal solution algorithm based on DE-PSO is designed to solve the optimization equation, and finally the spatial coordinate position where the partial discharge signal is generated is located.

[0243] Compared with the prior art, the present invention provides a method for denoising and positioning partial discharge signals of a switchgear cabinet combining an improved wavelet threshold algorithm and a time difference positioning method, and has the following beneficial effects: By adopting the improved wavelet threshold algorithm, the problems of over-denoising and incomplete denoising existing in the traditional denoising algorithm can be effectively overcome, the effective separation of signals and noise is realized, and the performance of the denoising algorithm is improved; The generalized cross-correlation time delay estimation algorithm based on PHAT-SCOT joint weighting is designed, which can effectively overcome the influence of complex on-site environment and various interferences, improve the accuracy of time delay estimation, and thus improve the positioning accuracy of partial discharge signals; A continuous adaptive wavelet threshold function is constructed, and by adjusting the adjustment parameter k of each wavelet decomposition layer, the characteristics of the threshold function are adaptively adjusted, which can better adapt to the denoising of partial discharge signals under different noise conditions and improve the denoising effect; The local discharge optimal solution algorithm based on DE-PSO is designed, which can effectively solve the defects existing in the existing algorithms in terms of optimization algorithm parameter setting and threshold selection, and improve the positioning accuracy and efficiency.

[0244] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A switch cabinet partial discharge signal positioning method based on time difference positioning method, characterized in that: The following steps are involved: Step 1: Collect signals using an ultrasonic partial discharge signal acquisition circuit, and filter and amplify the signals; Step 2, determining the wavelet basis function by comparing the denoising effects of different wavelet basis functions and different decomposition layers; Step 3: Calculate the adjustment parameter k of each wavelet decomposition layer and determine the threshold function corresponding to the wavelet coefficients of each layer; Step 4: For a given threshold λ0, calculate the unbiased likelihood estimate, and then minimize the threshold to obtain the optimal threshold of each wavelet decomposition layer; Step 5: Substitute the adjustment parameter k into the threshold function, determine the threshold function of each layer, and process the wavelet coefficients to obtain the estimated value of the wavelet coefficients; Step 6: Perform wavelet inverse reconstruction on the processed wavelet coefficients to obtain a denoised partial discharge signal; Step 7: Establish a switch cabinet partial discharge location equation group; Step 8, using a generalized cross-correlation delay estimation algorithm based on phase transform-smoothed coherence transform (PHAT-SCOT) joint weighting to estimate the delay; Step 9, substituting the estimated value of the time delay into the established switch cabinet partial discharge location equation group to obtain the optimization equation; Step 10: Solve the optimization equation using the partial discharge optimal solution algorithm based on the differential evolution algorithm and the particle swarm optimization algorithm (DE-PSO) to obtain the spatial coordinate position of the partial discharge signal.

2. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1 is characterized in that: The ultrasonic partial discharge signal acquisition circuit in step 1 includes: an autotransformer, a step-up transformer, a protection resistor, a coupling capacitor, a measurement impedance, a switch cabinet, a discharge model, an ultrasonic sensor, a preamplifier and an oscilloscope, wherein the preamplifier includes a filter and a signal amplifier; The step 1 comprises: Step 102: filtering the collected signal through the filter to remove high-frequency noise; Step 103: amplify the filtered signal by the signal amplifier to increase the signal strength.

3. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1 is characterized in that: The threshold function in step 3 is a continuous adaptive wavelet threshold function, which adaptively adjusts the characteristics of the threshold function according to the characteristics of the partial discharge signal at different wavelet decomposition layers to adapt to the partial discharge signal denoising under different noise conditions. The continuous adaptive wavelet threshold function is expressed as the following formula: In the formula, w x,k is the wavelet coefficient after threshold function processing, x is the wavelet coefficient, λ is the threshold, and k is the adjustment parameter of the threshold function.

4. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 3 is characterized in that: The value of the adjustment parameter k in the continuous adaptive wavelet threshold function is used to change the variation trend of the threshold function.

5. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 4 is characterized in that: For different partial discharge signal decompositions, the threshold function of each signal decomposition layer is determined by adjusting the parameter k to adapt to the characteristics of each partial discharge signal and the proportion of noise energy in each decomposition layer.

6. A switch cabinet partial discharge signal positioning method based on the time difference positioning method according to claim 5, characterized in that: The adjustment parameter k is determined according to the ratio of signal to noise, and its calculation formula is: In the formula, E di is the energy of the noisy signal in the i-th layer of wavelet decomposition, E ni is the energy of the noise in the i-th layer of wavelet decomposition, E di and E ni Calculated based on the scale factor and detail factor; Assume that when the i-th layer signal is all noise, E ni ≈E di , in this case k is 1; when the i-th layer signal contains no noise at all, E ni =0, k is 0; 0≤k≤1.

7. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1 is characterized in that: The step 7 is specifically as follows: Assume that S(x,y,z) is the coordinate of the discharge source inside the switch cabinet, and place four ultrasonic sensors for partial discharge detection. The sensor positions are P1(x1,y1,z1), P2(x2,y2,z2), P3(x3,y3,z3), and P4(x4,y4,z4). Construct a set of three-variable quadratic equations to solve the position of the discharge source S. Assuming that the sound speed v represents the ultrasonic propagation speed, the arrival time of the ultrasonic signal is t1, t2, t3, and t4, the method of constructing the optimization problem is used for indirect solution, and the objective function L is expressed as the following formula: Where, L≥0, if and only if L=0, x, y, z are solutions to the system of equations; The constructed optimization problem is expressed as the following formula: The above optimization problem is solved using the optimal solution algorithm to find the optimal solution of the x, y, and z values ​​that meet the requirements to obtain the partial discharge source coordinates S (x, y, z) and realize the localization of the partial discharge source.

8. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1 is characterized in that: The step 8 comprises the following steps: Step 801: Perform cross-correlation operation on signal x1(t) and signal x2(t) to obtain R 12 (τ); Step 802: The cross-correlation result R 12 (τ) Generalized weighting is performed using the PHAT-SCOT joint weighting factor to obtain the power spectral density function G 12 (ω); Step 803: The power spectrum density function G 12 (ω) is subjected to inverse Fourier transform and peak detection to obtain the delay estimation value.

9. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1, characterized in that: The step 9 is specifically as follows: Assume that the mathematical model of signals x1(t) and x2(t) is expressed as follows: Where: x1(t), x2(t) are delay estimation signals; s(t) is the signal source; m1(t), m2(t) are the white noises superimposed on the delay estimation signals; T is the time delay between the above delay estimation signals; The cross-correlation function R of the above two delay estimation signals is 12 (τ) is expressed as: R 12 (τ)=E[x1(t)x2(t-τ)] The cross-correlation function R of the signals x1(t) and x2(t) 12 (τ) and its cross power spectral density function G 12 The relational expression of (ω) is: The generalized cross-correlation function is expressed as: Where: ψ(ω) is the weighting function; The PHAT weighting factor expression is as follows: Where: is the cross power spectrum of signals x1(t) and x2(t); The SCOT weighting factor expression is as follows: Where: is the autopower spectrum of the signal x1(t); is the autopower spectrum of the signal x2(t); The expression of PHAT-SCOT joint weighting factor is as follows: Where: is the cross-power spectrum of the noise received by the signals x1(t) and x2(t); ρ is the correlation factor.

10. The method for locating partial discharge signals of a switch cabinet based on the time difference positioning method according to claim 1, characterized in that: The step 10 comprises the following steps: Step 1001: Initialize two parallel iterative sub-populations, DE and PSO, and use the DE algorithm to find the global optimal solution best. DE , find the global optimal solution best through PSO algorithm PSO ; Step 1002: Compare best DE and best PSO The speed formula of the PSO algorithm is updated based on the size of the two optimal solutions, and the iteration is stopped after the maximum number of iterations is reached. Step 1003: Compare the global optimal solution of the DE algorithm and the global optimal solution of the PSO algorithm generated in each iteration, and take the smallest one of the two as the global optimal solution of the DE-PSO algorithm.

Images