Fault source positioning method and system based on Wiener iterative filtering

Abstract

The invention provides a fault source positioning method and system based on Wiener iterative filtering. The method comprises the following steps: realizing synchronous acquisition of noise-containing partial discharge signals through a multi-channel sensor array arranged at two ends of a stator bar or along the line; performing adaptive noise reduction on the signal by using an iterative Wiener filter so as to highlight a discharge characteristic; estimating the relative time delay of the discharge pulse reaching different sensors according to the filtered signal; and based on the time delay, determining the position of the defect in the stator bar through a time difference positioning mechanism. According to the method, high-intensity background noise mixed in test signals can be effectively filtered out, the recognition degree of discharge characteristic signals is remarkably enhanced through optimization iteration of the weight of the self-adaptive filter, and high-precision judgment of the position of a defect source is achieved in combination with a time difference positioning mechanism. The method is not only suitable for partial discharge signal processing in power equipment, but also provides an efficient and stable signal processing tool for identifying an abnormal discharge event in a time domain dielectric response signal.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of discharge positioning, and particularly relates to a fault source positioning method and system based on Wiener iterative filtering BACKGROUND

[0002] Under the promotion of the "double carbon" goal, as the core equipment of clean energy, the stator winding main insulation of hydroelectric generating set is subjected to long-term electrical-thermal-mechanical multiple stresses, and insulation aging and partial discharge become the primary cause of unit failure (accounting for more than 50% of accident statistics). At present, non-destructive testing such as insulation resistance, dielectric loss or partial discharge quantity is generally used in engineering to evaluate the insulation state, but these means can only reflect the overall deterioration trend and are difficult to accurately identify and locate defects in the early stage. The traditional partial discharge positioning technology (amplitude method, time difference method, electrical mutation method) is easily disturbed by noise in the complex electromagnetic environment on site, and has poor robustness and insufficient reliability in the scene of multi-source discharge or strong background noise, which is difficult to meet the needs of rapid and accurate maintenance.

[0003] The core pain points of the prior art are that: complex on-site noise seriously reduces the positioning accuracy of traditional partial discharge, leading to misjudgment or missed judgment; non-destructive testing and conventional positioning methods cannot simultaneously achieve "determination of existence or non-existence + accurate positioning" of early insulation defects; existing algorithms have large calculation amount and high hardware dependence, and are difficult to be deployed in real time on site of a hydropower station. Therefore, there is an urgent need for a new partial discharge positioning method with strong anti-interference ability, low calculation complexity, high positioning accuracy and easy engineering implementation. SUMMARY

[0004] The purpose of the application is to solve the technical problems in the above background, and a fault source positioning method based on Wiener iterative filtering is proposed, which comprises the following steps:

[0005] S1. Synchronously collecting noisy partial discharge signals through a multi-channel sensor array arranged at both ends of the stator bar or along the line;

[0006] S2. Using an iterative Wiener filter to adaptively denoise the signals to highlight the discharge characteristics;

[0007] S3. Estimating the relative time delay of the discharge pulse arriving at different sensors according to the filtered signals;

[0008] S4. Based on the time delay, determining the position of the defect in the stator bar through a time difference positioning mechanism.

[0009] In a preferred scheme, the iterative Wiener filter updates the filter weights adaptively by online estimating the input / output noise variance ratio.

[0010] In a preferred scheme, the iterative update step is adaptively adjusted according to the energy of the gradient vector to avoid matrix inversion operation.

[0011] In the preferred scheme, a forgetting factor is introduced during the filter weight update process to track the slow changes in the signal statistical properties.

[0012] In the preferred scheme, the forgetting factor ranges from 0.9 to 1.

[0013] In the preferred embodiment, the time-difference positioning mechanism utilizes the known propagation speed of the discharge pulse in the stator bars to convert the time delay into distance.

[0014] In the preferred embodiment, when the change in defect location obtained from multiple consecutive iterations is less than a set threshold, the iteration is terminated and the final location result is output.

[0015] A system for implementing a fault source localization method based on Wiener iterative filtering, comprising:

[0016] A multi-channel high-speed acquisition module is used to synchronously acquire partial discharge signals of stator bars;

[0017] The real-time signal preprocessing module is used to perform DC removal, normalization, and filtering on the acquired signals;

[0018] The iterative Wiener filter calculation module is used to perform weight updates and time delay estimation.

[0019] The positioning calculation and output module is used to calculate the defect location based on the time delay and output the positioning result.

[0020] The communication interface module is used to upload the positioning results to a remote monitoring center or mobile terminal.

[0021] In the preferred embodiment, the iterative Wiener filter calculation module is implemented based on FPGA, GPU or embedded multi-core DSP to meet real-time requirements.

[0022] The preferred solution also includes:

[0023] The adaptive threshold alarm module triggers an alarm when the rate of change of the defect location exceeds a set threshold.

[0024] The data storage and backtracking module is used to store historical location data and support fault development trend analysis.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] (1) By minimizing the disturbance between the input and output signals, the present invention provides a better estimate of the input noise, thereby achieving suppression and unbiased estimation of the adaptive filter parameters.

[0027] (2) The present invention iteratively updates the weight coefficients w(j) based on the Wiener filtering algorithm, thereby avoiding the matrix inversion operation and reducing the high computational complexity of the algorithm, making the algorithm easy to implement; at the same time, the estimation of arrival time difference is adopted to make a more accurate estimate of the fault location source. Attached Figure Description

[0028] Figure 1 This is a comparison diagram of square wave fault source signal localization methods.

[0029] Figure 2 This is a comparison chart of signal location methods for real PD signal fault sources. Detailed Implementation

[0030] Example 1

[0031] This embodiment provides a fault source localization method based on Wiener iterative filtering, the steps of which are as follows:

[0032] A. Noisy input vector x i (j),x i (j-1)...,x i (j-M+1)

[0033] At time j, the signal emitted by the fault source is sampled. After processing the sampled signal, the filter input vector x(j) at the current time j is obtained. The noisy filter input vector xj at the current time j is then calculated from x(j) and υ(j). i (j); The current time j and the M noisy inputs x from the previous M-1 times are combined. i (j),x i (j-1)...,x i (j-M+1) constitutes the noisy input x of the filter at the current time j. i (j), x i (j)=[x i (j),x i (j-1)...,x i (j-M+1)] T Where M is the order of the filter, and the superscript T represents the transpose. Where x i (j)=x(j)+υ(j).

[0034] B. Generation of weight vectors

[0035] The filter generates weight coefficients w(j), w(j-1), ..., w(j-M+1) for the current time j and the previous M-1 times, and constructs the weight coefficient vector w(j) for the current time j from these M weight coefficients, where w(j) = [w(j), w(j-1), ..., w(j-M+1)]. TAt the current time j=0, w(j)=0.

[0036] C. The autocorrelation matrix Ψ(j) containing the noisy input vector, the cross-correlation vector v(j) containing the noisy input vector and the desired response, and the autocorrelation term r of the desired response are calculated as follows:

[0037] The filter generates the current time j, and Ψ(j) and ν(j) are calculated as follows:

[0038]

[0039] and

[0040]

[0041] In the formula, λ represents the forgetting factor, which ranges from 0.9 to 1.

[0042] D. Calculate the ratio of output noise variance to input noise variance and the gradient vector.

[0043] This is the ratio of the output noise variance to the input noise, where... and These are the output noise variance and the input noise variance, respectively. This is the gradient vector.

[0044] E. Step size for calculating weight vector updates

[0045] This is the step size for updating the weight vector.

[0046] Where a = w T gg T Ψg-w T Ψgg T gg+gg T gg T ν

[0047] b = 2g T gw T ν-g T gw T Ψw-g T gζ+g T Ψgw T w+g T Ψgr

[0048] c = w T Ψgw T w+w T Ψgr-g T νw T wg T νr-w T gwT Ψw T +2w T gw T ν-w T gζ

[0049] F. Calculate and update the weight vector obtained from steps B, D, F, and G.

[0050] The filter calculates the weight vector w(j+1) for the next time step j+1, w(j+1) = w(j) + μ(j)g(j).

[0051] G. Calculate the positioning position based on step G.

[0052] If w m+1 If (j)≥0, the arrival time delay estimate τ(j) is calculated as follows:

[0053]

[0054] If w m+1 The arrival time delay estimate τ(j) is calculated as follows: (j) < 0.

[0055]

[0056] Among them, w m+1 (j) refers to the maximum value of the weight coefficient vector w(j) at the (m+1)th position, and m(j) refers to the delay estimate of j at the current time.

[0057] H, Iteration

[0058] Let j = j + 1, and repeat steps A, B, C, D, E, F, G, and H until the algorithm ends.

[0059] Example 2

[0060] Simulation experiment:

[0061] To verify the effectiveness of the present invention, a comparative simulation experiment was conducted and compared with the methods of methods 1 and 2.

[0062] The order M of the filter in the simulation experiment is set to 100. Figure 1 ), 100 ( Figure 2 ).

[0063] exist Figure 1 In this method, a square wave fault source is used as the fault signal, and the fault signal is set at 300m (the actual location). The algorithm parameters for Method 1 are set to μ (step size) = 0.0001; the parameters for Method 2 are set to μ = 0.0001; and the parameters for this invention are set to λ (forgetting factor) = 0.9999.

[0064] existFigure 2 The fault signal uses a real PD signal fault source, and the fault signal is set at 300m (the actual location). The algorithm parameters for Method 1 are set to μ = 0.0001; the parameters for Method 2 are set to μ = 0.0001; and the parameters for this invention are set to λ = 0.9999.

[0065] Depend on Figure 1 , Figure 2 It can be seen that the method of the present invention has higher positioning accuracy.

[0066] Table 1 below summarizes the positioning results of Method 1, Method 2 and the present invention in two different fault noise environments. As can be seen from Table 1, the method of the present invention can achieve high-precision positioning for different fault noise sources.

[0067] Table 1 Comparison of accuracy of different methods

[0068]

[0069] This invention discloses a method for locating discharge fault sources in the stator winding of a large hydro-generator unit based on Wiener iterative filtering. The main steps are: A) Generation of a noisy input vector: At the current time j, the signal generated by the signal source is sampled. After processing the sampled signal, the filter input vector x(j) at the current time j is obtained. The noisy filter input vector xj at the current time j is then calculated from x(j) and υ(j). i (j); B. Generation of weight vector: the filter generates w(j) for the current time j and the previous M-1 times, w(j-1)…w(j-M+1); C. Calculation of Ψ(j) and v(j); D. Calculation of the ratio of output noise variance to input noise variance and gradient vector; E. Calculation of the step size for weight vector update; F. Weight vector calculation and update; G. Calculation of localization position; H. Iteration: let j = j+1 until noise control ends. From Figure 1 , Figure 2 It can be found that the method of the present invention has higher positioning accuracy and can achieve high-precision positioning for various fault sources.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault source localization method based on Wiener iterative filtering, characterized by: Includes the following steps: S1. Synchronous acquisition of noisy partial discharge signals is achieved by arranging a multi-channel sensor array at both ends or along the stator bar; S2. Adaptive noise reduction of the signal is performed using an iterative Wiener filter to highlight the discharge characteristics; S3. Estimate the relative time delay of the discharge pulse reaching different sensors based on the filtered signal; S4. Based on the time delay, determine the location of the defect in the stator bar through a time difference positioning mechanism.

2. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: The iterative Wiener filter adaptively updates the filter weights by estimating the input / output noise variance ratio online.

3. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: The iterative update step size is adaptively adjusted based on the energy of the gradient vector to avoid matrix inversion operations.

4. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: A forgetting factor is introduced during the filter weight update process to track the slow changes in the signal's statistical properties.

5. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: The forgetting factor ranges from 0.9 to 1.

6. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: The time-difference positioning mechanism uses the known propagation speed of the discharge pulse in the stator bars to convert the time delay into distance.

7. The fault source localization method based on Wiener iterative filtering according to claim 1, characterized in that: When the change in defect location obtained from multiple consecutive iterations is less than a set threshold, the iteration is terminated and the final location result is output.

8. A system for implementing the fault source localization method based on Wiener iterative filtering according to any one of claims 1 to 6, characterized in that it comprises: A multi-channel high-speed acquisition module is used to synchronously acquire partial discharge signals of stator bars; The real-time signal preprocessing module is used to perform DC removal, normalization, and filtering on the acquired signals; The iterative Wiener filter calculation module is used to perform weight updates and time delay estimation. The positioning calculation and output module is used to calculate the defect location based on the time delay and output the positioning result. The communication interface module is used to upload the positioning results to a remote monitoring center or mobile terminal.

9. The system according to claim 1, characterized in that: The iterative Wiener filter calculation module is implemented based on FPGA, GPU or embedded multi-core DSP to meet real-time requirements.

10. The system according to claim 1, characterized in that: it also... include: The adaptive threshold alarm module triggers an alarm when the rate of change of the defect location exceeds a set threshold. The data storage and backtracking module is used to store historical location data and support fault development trend analysis.

Images