Traction transformer built-in optical fiber partial discharge positioning method

Abstract

The invention belongs to the technical field of power equipment state monitoring, and relates to a traction transformer built-in optical fiber partial discharge positioning method. According to the method, a plurality of Fabry-Perot optical fiber ultrasonic sensors are arranged in a transformer to collect partial discharge acoustic signals, multichannel data are obtained in a synchronous sampling mode, and the time difference of arrival of the signals is extracted based on a generalized cross-correlation phase transformation method. Dynamic sound velocity compensation is realized by constructing a second-order fitting model of sound velocity and temperature in oil, an initial positioning result is obtained in combination with a Chan algorithm, and three-dimensional coordinates of a partial discharge source are further obtained through nonlinear optimization iteration. Data screening is carried out through characteristic parameters such as a signal-to-noise ratio, a peak rise time ratio and kurtosis, and the positioning reliability is improved. The method has the characteristics of strong anti-electromagnetic interference capability, suitability for the internal working condition of the oil-immersed traction transformer and high positioning precision, and can be used for online partial discharge monitoring and fault early warning of traction power supply equipment.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, specifically to a method for locating partial discharge using an embedded optical fiber in a traction transformer. Background Technology

[0002] Traction transformers are the core equipment of rail transit power supply systems, and their operational stability directly determines the reliability of rail transit power supply.

[0003] Partial discharge is the main cause of insulation deterioration in traction transformers. If it cannot be detected and located in time, it can easily lead to equipment failure or even power outages.

[0004] Currently, partial discharge detection and localization technologies mainly include pulse current method, ultra-high frequency method, and ultrasonic method.

[0005] Among them, the ultrasonic method is widely used in the field due to its strong anti-interference ability and high positioning accuracy.

[0006] Existing ultrasonic positioning systems mostly adopt external sensor layouts, such as a four-channel high-speed sampling system based on MEMS fiber optic ultrasonic sensors. By arranging a sensor array outside the device, partial discharge positioning is achieved using the time difference of arrival (TDOA) algorithm.

[0007] However, the existing technology has obvious drawbacks: First, external sensors are easily affected by environmental noise, resulting in significant signal attenuation and positioning accuracy being greatly affected by the obstruction of the external structure of the device.

[0008] Secondly, ultrasonic signals from high-frequency partial discharge areas inside the traction transformer (such as winding ends and insulation joints) are easily blocked by components such as the iron core and oil tank when transmitted outward, resulting in incomplete signal capture by external sensors.

[0009] Third, the existing system is not optimized for special operating conditions such as the internal oil medium, high temperature, and high pressure of the traction transformer, and the positioning algorithm does not consider the influence of the internal medium on the speed of sound, resulting in a large positioning deviation.

[0010] Fourth, the external layout is difficult to adapt to the complex internal structure of the traction transformer, making it impossible to accurately locate the partial discharge source.

[0011] Therefore, a partial discharge positioning method that is adapted to the internal working conditions of traction transformers, has high positioning accuracy, and stable signal capture is needed to address the shortcomings of existing external positioning technologies. Summary of the Invention

[0012] The present invention aims to provide a partial discharge positioning system and method suitable for the internal structure of traction transformers, which solves the problems of signal attenuation, large interference and low accuracy of existing external positioning technology, and realizes high-precision three-dimensional positioning of partial discharge sources.

[0013] A method for locating partial discharge using an embedded optical fiber in a traction transformer includes the following steps:

[0014] The first step is the deployment and signal acquisition of the built-in fiber optic sensor:

[0015] Four miniature MEMS Fabry-Perot fiber ultrasonic sensors were selected, and after being packaged in a ceramic + polytetrafluoroethylene composite, they were deployed in a tetrahedral structure in the high-occurrence area of ​​partial discharge inside the traction transformer.

[0016] A signal link consisting of "sensor - fiber optic link - photoelectric conversion module - high-speed acquisition card" was constructed to simultaneously acquire partial discharge ultrasonic signals received by four sensors. in Number the sensors;

[0017] The second step is to calibrate the sensor coordinates and sound velocity parameters.

[0018] The calibration coordinates of the i-th sensor are The coordinates of the partial discharge power source are ;

[0019] The sound velocity of transformer oil at different temperatures was measured, and a "temperature-sound velocity" fitting model was established to obtain the dynamically corrected sound velocity c.

[0020]

[0021] in The sound velocity is obtained through fitting and corrected based on real-time temperature input during operation.

[0022] Step 3: Extraction of partial discharge signal time difference:

[0023] Using the first sensor as a reference, the arrival time difference between the signals from the other sensors and the reference signal is obtained using the cross-correlation method. ( ),in yes and The time difference of arrival;

[0024] Step 4: Solving for the three-dimensional coordinates of the partial discharge source:

[0025] Based on sensor coordinates, dynamic sound velocity c(T), and extracted time difference of arrival. The Chan algorithm based on TDOA is used for the initial localization of the local discharge source.

[0026] Chan's algorithm transforms the nonlinear TDOA positioning equation into a linear least squares form, and its mathematical model is as follows:

[0027] ,

[0028] in

[0029] Transforming the above equation into a matrix, we get:

[0030] The closed-form solution for the initial coordinates of the partial discharge source is:

[0031] The initial localization results are then input into the Levenberg-Marquardt algorithm for iterative optimization to obtain the final three-dimensional localization coordinates of the partial discharge source. To improve positioning accuracy and reduce the impact of noise errors.

[0032] The original system of equations for TDOA is as follows:

[0033]

[0034]

[0035]

[0036] Step 5: Verify the location results:

[0037] Extracting ultrasound signals Waveform characteristic parameters (amplitude-rise time ratio) , cliff ), remove abnormal data, output positioning coordinates and deviation values, and ensure that the positioning deviation is ≤5cm.

[0038] The amplitude-rise time ratio : for ultrasound signals The ratio of the peak value to the rise time of the signal, expressed in V / s;

[0039] The kurtosis The calculation formula is:

[0040] Step 6: Verification and output of positioning results:

[0041] The positioning results are corrected by the data processing module based on the threshold determined by experimental calibration or statistical confidence interval.

[0042] The spatial location of the local source is visualized on a dedicated platform, and the positioning coordinates, deviation value and confidence level are output to ensure that the positioning deviation is ≤5cm. Attached Figure Description

[0043] To more clearly illustrate the method of the present invention, the accompanying drawings used in the present invention will be described below.

[0044] Figure 1 This is a diagram of the tetrahedral deployment structure of the built-in fiber optic sensor in the traction transformer.

[0045] Figure 2 This is a flowchart of the partial discharge localization method using built-in optical fiber in a traction transformer. Detailed Implementation

[0046] This embodiment provides a specific implementation of the steps described above in the invention:

[0047] 1. Sensor deployment:

[0048] A MEMS Fabry-Perot fiber optic ultrasonic sensor with a diameter ≤5mm and a length ≤20mm was selected, and the insulation resistance after packaging was ≥1000MΩ.

[0049] Four sensors are fixed to the ends of the windings and the corresponding areas of the insulating joints inside the traction transformer in a regular tetrahedral structure. The sensor spacing is 0.5m~1m and the optical fiber bending radius is ≥5cm.

[0050] 2. Signal Acquisition:

[0051] A 4-channel high-speed acquisition card with a sampling frequency ≥1MHz was used, and a photoelectric detection module that met the system's sensitivity and signal-to-noise ratio requirements was selected. The response and noise performance within the target frequency band (e.g., 20 kHz–500 kHz) needed to meet the experimentally calibrated minimum detectable sound pressure level (MDP) requirement, and sensor signals were acquired synchronously. ;

[0052] 3. Sound velocity calibration:

[0053] Within a temperature range of -40℃ to 80℃, the sound velocity of transformer oil was measured, and a fitting model was established. (T represents oil temperature)

[0054] 4. Time difference extraction:

[0055] Calculating signals using the cross-correlation method and Time difference Δ , and Time difference Δ , and Time difference Δ ;

[0056] 5. Coordinate Determination:

[0057] Based on sensor calibration coordinates And dynamically correct the speed of sound c(T), and construct a distance relationship model based on TDOA:

[0058] , .

[0059] in:

[0060] The above nonlinear equations are transformed into matrix form through linearization:

[0061] in Let be the three-dimensional coordinate vector of the partial discharge source to be determined. From this, a closed-form analytical solution can be obtained:

[0062] in It serves as the initial solution for subsequent iterative optimization.

[0063] Based on the initial positioning, in order to improve noise resistance and positioning accuracy, the Levenberg–Marquardt (LM) algorithm is further introduced for nonlinear least squares fitting.

[0064] Its iterative model is:

[0065] in, Let Jacobian matrix be the error function. Let be the residual vector of the k-th iteration. The damping factor is adaptively adjusted to switch between gradient descent and Gauss-Newton methods to obtain more robust convergence performance.

[0066] The iteration termination condition is: It may reach the maximum number of iterations.

[0067] Finally, the precise coordinates of the partial discharge were obtained.

[0068] 6. Result Verification:

[0069] Extract signal , Remove <10 5 mV / μs or For abnormal data <3, output a result with a positioning deviation ≤5cm.

Claims

1. A method for locating partial discharge using an embedded optical fiber in a traction transformer, characterized in that, The method includes the following steps: a) Deployment and signal acquisition of built-in fiber optic sensors: Four miniature MEMS Fabry-Perot fiber optic ultrasonic sensors were selected and deployed in a tetrahedral structure in the high-frequency partial discharge area inside the traction transformer after being encapsulated in a ceramic + polytetrafluoroethylene composite package. A signal link of "sensor-fiber optic link-photoelectric conversion module-high-speed acquisition card" was constructed to simultaneously acquire the partial discharge ultrasonic signals received by the four sensors. Where i = 1, 2, 3, 4 are the sensor numbers; b) Sensor coordinate and sound velocity parameter calibration: The calibration coordinates of the i-th sensor are as follows: The coordinates of the partial discharge power source are The sound velocity of transformer oil at different temperatures was measured, and a "temperature-sound velocity" fitting model was established to obtain the dynamically corrected sound velocity c. in The velocity of sound is obtained through fitting, and is corrected based on real-time temperature input during operation. c) Partial discharge signal time difference extraction: Using the first sensor as a reference, the arrival time difference between the signals from the other sensors and the reference signal is obtained using the cross-correlation method. (i=2,3,4), where For signal and The time difference of arrival; d) Solving for the three-dimensional coordinates of the partial discharge source: based on sensor coordinates, dynamic sound velocity c(T), and extracted time difference of arrival. The Chan algorithm, based on TDOA, is used for the initial localization of the local discharge source. The Chan algorithm transforms the nonlinear TDOA localization equation into a linear least squares form, and its mathematical model is as follows: ; in: ; Transforming the above equation into a matrix, we get: ; The closed-form solution for the initial coordinates of the partial discharge source is: The initial localization results are then input into the Levenberg-Marquardt algorithm for iterative optimization to obtain the final three-dimensional localization coordinates of the partial discharge source. To improve positioning accuracy and reduce the impact of noise errors; e) Verification of localization results: Extraction of ultrasound signals The waveform characteristic parameters (amplitude-rise time ratio PRRᵢ, kurtosis Kuᵢ) are used to remove abnormal data and output the positioning coordinates and deviation values ​​to ensure that the positioning deviation is ≤5cm. The amplitude-rise time ratio PRRᵢ is the ratio of the peak value of the ultrasonic signal sᵢ to the rise time of the signal, in units of V / s. The formula for calculating the kurtosis Kuᵢ is as follows: ; f) Verification and output of positioning results: The positioning results are corrected for errors through the data processing module, based on the threshold determined by experimental calibration or statistical confidence interval; the spatial location of the partial discharge source is visualized on a dedicated platform, and the positioning coordinates, deviation value and confidence level are output to ensure that the positioning deviation is ≤5cm.

2. The method according to claim 1, characterized in that, The four sensors are arranged in a tetrahedral structure with a spacing of 1.5m to 2m to ensure effective positioning of the local discharge power source.

3. The method according to claim 1, characterized in that, The fiber optic sensor uses a MEMS Fabry-Perot ultrasonic sensor, which features high precision, low noise, and electromagnetic interference resistance.

4. The method according to claim 1, characterized in that, In step c), the signal arrival time difference between each sensor and the reference sensor is calculated using the cross-correlation method, and the preliminary position of the partial discharge source is calculated accordingly.

5. The method according to claim 1, characterized in that, In step d), the Chan algorithm is used to calculate the preliminary positioning results, and the three-dimensional coordinates of the local discharge source are corrected by combining the dynamic sound velocity temperature compensation model.

6. The method according to claim 1, characterized in that, In step e), the Levenberg–Marquardt (LM) algorithm is used to optimize the preliminary positioning results. By minimizing the residuals, the positioning accuracy is improved and the positioning error is controlled within 5 cm.

7. The method according to claim 1, characterized in that, The three-dimensional coordinates of the partial discharge source include coordinates in the x, y, and z axes, and can be updated in real time as part of the transformer's internal condition monitoring.

Images