Layered detection method and system for oil temperature of transformer based on thin-plate spline basis function

Abstract

The invention discloses a transformer oil temperature layered detection method and system based on a thin plate spline primary function. The method comprises the following steps: forming a detection point attribute by using a propagation time average value, a fusion amplitude, a fusion speed and a fusion coordinate corresponding to each transducer; mapping attributes of each detection point into a three-dimensional coordinate system of the oil tank, and obtaining a detection sparse network based on a graph structure sparse method; dividing the oil tank into a plurality of equal-height layers along the height direction of the oil tank, and performing thin-plate spline difference fitting on the detection sparse network by taking the minimum bending energy as an optimization target based on the thin-plate spline primary function of each layer to obtain a global detection network; determining a speed estimation value of the ultrasonic signal corresponding to any point in the oil tank by using the high-confidence representative value, the low-confidence representative value and a radial basis function of a thin plate spline basis so as to establish a sound velocity distribution field in the oil tank; and according to the sound velocity distribution field in the oil tank, a Bayesian statistical method is adopted, and a corrected oil temperature estimation value is obtained based on a Gaussian likelihood oil temperature observation model.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of electrical equipment maintenance, and particularly relates to a transformer oil temperature layered detection method and system based on a thin plate spline base function. BACKGROUND

[0002] Oil-immersed transformers play a crucial role in power systems, ensuring stability and security of power supply through effective voltage transformation and efficient energy transmission, while improving the overall efficiency of the power system. The temperature of the transformer plays a crucial role in its operation and maintenance, affecting its performance, efficiency, lifespan, and safety. The insulation materials inside the transformer, such as transformer oil and insulating paper, are highly sensitive to temperature. High temperatures can accelerate the aging and degradation of insulation materials, thereby shortening the service life of the transformer. Generally, the lifespan of insulation materials decreases exponentially with temperature. For example, for every 8℃ increase, the lifespan of insulation materials may be halved. The load capacity of the transformer is directly limited by its temperature. When the load of the transformer increases, the temperature also rises. If the temperature exceeds the rated value, it will cause the transformer to overheat, affecting its normal operation. The design of the transformer usually contains a temperature limit to ensure safe and efficient operation within this limit. Through effective temperature control, the load capacity of the transformer can be improved to meet the needs of the power system. In addition, the efficiency and loss of the transformer are closely related to temperature. Temperature rise will increase copper loss and iron loss, thereby reducing the efficiency of the transformer. Efficient transformer operation relies on effective temperature management to reduce unnecessary energy loss and improve overall operating efficiency. Especially in high-load operation, temperature management is particularly important. Moreover, due to the spatial structure of the transformer, there are data detection points that cannot be measured, resulting in inaccurate and unreliable three-dimensional temperature field reconstruction. Traditional point measurement cannot fully cover the temperature field of the oil tank, and cannot accurately capture the temperature abnormal heating area in the traditional monitoring blind area such as the oil flow stagnation area. SUMMARY

[0003] To solve the problems in the prior art, the application provides a transformer oil temperature layered detection method and system based on a thin plate spline base function, which realizes the reconstruction of the three-dimensional temperature field of the oil tank based on the graph structure sparse method and the thin plate spline base function, can fit the data detection points in the three-dimensional temperature field that cannot be measured due to spatial structure factors, and provides strong support for maintenance personnel to detect the temperature of the transformer and locate faults, thereby significantly improving the maintenance efficiency.

[0004] The application adopts the following technical solutions.

[0005] This invention proposes a layered detection method for transformer oil temperature based on thin-plate spline basis functions. Multiple transducers are arranged on the surface of the transformer tank, each transducer receiving ultrasonic signals directed along different paths; including: The propagation time, amplitude, and velocity of ultrasonic signals on different paths corresponding to each transducer are obtained, and the coordinates of the geometric centroids of all path intersections corresponding to each transducer are obtained. The amplitude, velocity, and coordinates are fused based on the propagation time, and the average propagation time, fused amplitude, fused velocity, and fused coordinates corresponding to each transducer are used to construct the detection point attributes. The attributes of each detection point are mapped to the three-dimensional coordinate system of the fuel tank, and a detection sparse network is obtained based on the graph structure sparsity method. The fuel tank is divided into multiple layers of equal height along its height, and thin-plate spline basis functions are established for each layer. Based on the thin-plate spline basis functions of each layer, with the minimum bending energy of the thin plate as the optimization objective, thin-plate spline difference fitting is performed on the detection sparse network to obtain the global detection network. Using the high-confidence representative value and low-confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network, and the radial basis function of the thin-plate spline basis, the velocity estimate of the ultrasonic signal corresponding to any point in the fuel tank is determined to establish the sound velocity distribution field in the fuel tank. Based on the sound velocity distribution field in the fuel tank, the corrected fuel temperature estimate is obtained by using Bayesian statistical methods and a Gaussian likelihood-based fuel temperature observation model.

[0006] transducer Reflection point and transducer A path , , The number of transducers, , For transducers The number of corresponding reflection points; transducer Receive path The amplitude of the ultrasonic signal is shown in the following formula:

[0007] In the formula, , Paths The amplitude and propagation time of the ultrasonic signal on the surface. For transducers The initial amplitude of the emitted ultrasonic signal, This is the overall attenuation factor; Based on transducer The propagation time of all received ultrasonic signals is fused with the amplitude of all ultrasonic signals, as shown in the following formula:

[0008] where, is the fusion amplitude of all the received ultrasonic signals. is the fusion amplitude of all the received ultrasonic signals.

[0009] According to the transmission angle of each transducer and the size of the transformer oil tank, the geometric length of each path is determined by geometric method. The speed of the ultrasonic signal on the corresponding path is as follows:

[0010] where, , is the speed of the ultrasonic signal on the path and the geometric length of the path, is the denoised ultrasonic time-domain signal, is the transmission pulse of the transducer, is the transmission period; The propagation time of all the received ultrasonic signals based on the transducer is fused to obtain the fusion propagation time of all the received ultrasonic signals based on the transducer The speed of all the received ultrasonic signals based on the transducer is fused to obtain the fusion speed of all the received ultrasonic signals based on the transducer

[0011] where, is the fusion speed of all the received ultrasonic signals based on the transducer .

[0012] The geometric barycenter coordinates of all the path intersection segments of the transducer are determined by geometric barycenter method. where, is the path intersection point; The geometric barycenter coordinates of all the path intersection segments of the transducer are fused based on the propagation time of all the received ultrasonic signals to obtain the fusion geometric barycenter coordinates of all the path intersection segments of the transducer .

[0013] where, is the fusion coordinate of the geometric barycenter of all the path intersection segments of the transducer .

[0014] The corresponding propagation time average , fusion amplitude , fusion speed , fusion coordinate of the transducer constitute the detection point attribute , and the detection point set is ; each detection point attribute is mapped to a three-dimensional coordinate system of the oil tank, and a detection sparse network is obtained based on a graph structure sparse method; Based on the transducer Corresponding to the fusion coordinate constraint, the thin plate spline basis function is used to perform thin plate spline interpolation fitting on the detection sparse network, taking the minimum bending energy of the thin plate as the optimization objective, to obtain a global detection network and determine the detection point attribute obtained by the thin plate spline interpolation fitting , , The number of detection points obtained by the thin plate spline interpolation fitting.

[0015] Along the height direction of the oil tank, the oil tank is divided into multiple equal-height layers, and the height of each layer is not greater than the diameter of the effective coverage area of a single transducer transmitting signal; the center point coordinates of the first layer are , , The number of layers; the distance between any point in the layer and the center point is The geometric center point of the first layer is taken as the reference to construct the thin plate spline basis function of the first layer , as shown in the following formula:

[0016] The propagation matrix is constructed by the integral values of the thin plate spline basis functions of each layer on different paths , The number of transducers is , and the number of reflection points corresponding to one transducer is

[0017] Based on the propagation matrix, the corresponding fusion velocity of the transducer is used to establish a sound speed observation vector, as shown in the following formula:

[0018] In the formula, is the sound speed observation vector of the transducer with position coordinates , and is the fusion velocity corresponding to the transducer ; Based on the sound speed observation vector, the thin plate bending energy is established, as shown in the following formula:

[0019] In the formula, is the thin plate bending energy, is a second-order multiple index, is a second-order partial derivative. ​

[0020] The signal-to-noise ratio is calculated by using the fusion amplitude of each detection point in the global detection network, and the detection point with a signal-to-noise ratio not less than a set threshold is a high-confidence point, and the detection point with a signal-to-noise ratio less than the set threshold is a low-confidence point; The high-confidence interval is The low-confidence interval is , The number of high-confidence points is taken as the high-confidence representative value The median of the low-confidence interval is taken as the low-confidence representative value ; Each detection point in the global detection network is sorted in order of signal-to-noise ratio from small to large, and the ratio of the sorting sequence number of each detection point to + is taken as the confidence weight of each detection point, so as to establish the confidence weight diagonal matrix ; The weighted objective function of the high-confidence representative value and the low-confidence representative value is established, as shown in the following formula:

[0021] In the formula, is the weighted objective function of the high-confidence representative value and the low-confidence representative value, is the sound velocity observation vector of the high-confidence representative value, is the sound velocity observation vector of the low-confidence representative value, is a vector composed of the fusion velocities of all transducers, is a regularization term; The minimum change of the weighted objective function is taken as the optimization target to determine the high-confidence representative value and the low-confidence representative value.

[0022] The sound velocity estimation vector of any point in the oil tank is shown in the following formula:

[0023] In the formula, is the sound velocity estimation vector of any point in the oil tank, is the coordinate vector of any point in the oil tank, is the coordinate vector of the fitting barycenter point of the first layer of the oil tank, is a constant bias term, is a radial basis function of a thin plate spline basis, , is the distance between any point in the oil tank and the fitting barycenter point of the first layer of the oil tank, The value is 0.001; Using the propagation matrix and the sound velocity estimation vector at any point inside the tank, the velocity estimate of the ultrasonic signal at any point inside the tank is obtained, as shown in the following formula:

[0024] In the formula, This is the estimated velocity of the ultrasonic signal at any point inside the fuel tank. Based on the velocity estimate of the ultrasonic signal corresponding to any point inside the fuel tank, the sound velocity distribution field of the fuel tank is established.

[0025] The initial value of the predicted temperature is obtained by mapping the linear multivariate physical relationship between speed and oil temperature. Establish the Bayesian prior distribution of oil temperature. , , For based on The sound velocity fusion kernel matrix of thin plate spline basis. For smoothing coefficients, The prior covariance matrix; Based on the Bayesian prior distribution of oil temperature, the bending energy of the thin plate is... ; An oil temperature observation model based on Gaussian likelihood is established, as shown in the following equation: ,

[0026] In the formula, For the observed temperature measurement data, This represents the actual temperature field of the fuel tank. To measure noise, Here is the noise covariance matrix; The corrected oil temperature estimate is obtained by updating the posterior mean using closed-loop Bayesian algorithm, with the following relationship:

[0027] In the formula, This is the corrected oil temperature estimate.

[0028] This invention also proposes a layered detection system for transformer oil temperature based on thin-plate spline basis functions, comprising: The detection data processing module is used to obtain the propagation time, amplitude, and velocity of ultrasonic signals on different paths corresponding to each transducer, and to obtain the coordinates of the geometric centroids of all path intersections corresponding to each transducer; based on the propagation time, the amplitude, velocity, and coordinates are fused respectively, and the average propagation time, fused amplitude, fused velocity, and fused coordinates corresponding to each transducer are used to construct the detection point attributes; the attributes of each detection point are mapped to the three-dimensional coordinate system of the fuel tank, and a detection sparse network is obtained based on the graph structure sparsity method; An oil temperature estimation module is configured to divide the oil tank into multiple layers along the height direction of the oil tank, and to establish a thin plate spline base function for each layer; based on the thin plate spline base function for each layer, a thin plate spline interpolation fitting is performed on the detection sparse network with the minimum bending energy of the thin plate as an optimization target to obtain a global detection network; a high confidence representative value and a low confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network, and a radial basis function of the thin plate spline base are used to determine a velocity estimation value of an ultrasonic signal corresponding to any point in the oil tank to establish a sound velocity distribution field of the oil tank; and based on the sound velocity distribution field of the oil tank, a corrected oil temperature estimation value is obtained by using a Bayesian statistical method based on a Gaussian likelihood oil temperature observation model.

[0029] The application also provides a terminal, including a processor and a storage medium; the storage medium is used for storing instructions; and the processor is used for operating according to the instructions to execute the steps of the method.

[0030] The application also provides a computer readable storage medium, which stores a computer program; when the program is executed by a processor, the steps of the method are implemented.

[0031] Compared with the prior art, the application has at least the following beneficial effects: the application processes the pulse processing signal received by the transducer, and implements a layered temperature field reconstruction algorithm based on the graph structure sparse method and the thin plate spline base function of each layer, effectively solves the problem of missing temperature information of the traditional least square method in a complex oil flow environment, and improves the reconstruction accuracy of the three-dimensional temperature field; the application divides the oil tank into layers and uses the minimum bending energy optimization criterion based on the thin plate spline base function, which not only overcomes the lack of collaborative operation in single-path detection, but also successfully solves the problem of shielding interference of the ultrasonic wave caused by the internal metal components of the transformer; finally, the application uses the weighting factor to differentiate the high confidence points and the low confidence points, realizes adaptive measurement error suppression, and thus obtains more stable and reliable three-dimensional sound velocity field reconstruction results, so that the reconstruction results not only conform to the observation data, but also maintain physical rationality and spatial smoothness, and the temperature distribution of each layer is displayed in the form of a three-dimensional thermal diagram, which provides accurate and reliable decision basis for transformer condition maintenance.

[0032] The application realizes a transformer oil temperature layered ultrasonic detection technology based on a thin plate spline function, realizes all-around accurate monitoring of the internal temperature distribution of the transformer through layered detection of the transformer, and expands the sparse measurement points into a continuous three-dimensional temperature field distribution through a unique temperature field modeling method, thereby improving the accuracy and reliability of the transformer oil tank temperature detection. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A flowchart of the layered detection method for transformer oil temperature based on a thin plate spline base function according to the application is shown in the figure. Figure 2A three-dimensional temperature field main view of a transformer oil tank in the embodiment of the present application; Figure 3 A three-dimensional temperature field rotation view of a transformer in the embodiment of the present application; Figure 4 A top view of a temperature field of a transformer in the embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with the drawings in the embodiment of the present application. The embodiments described in the present application are only a part of the embodiments of the present application, but not all the embodiments. Based on the spirit of the present application, all other embodiments obtained by those skilled in the art without making creative efforts fall within the protection scope of the present application.

[0035] The present application proposes a layered detection method of transformer oil temperature based on thin plate spline basis function, as shown in Figure 1 The method comprises the following steps: Step 1, obtaining the propagation time, amplitude and speed of ultrasonic signals on different paths corresponding to each transducer, and obtaining the coordinates of the geometric center of intersection of all paths corresponding to each transducer; based on the propagation time, the amplitude, speed and coordinates are fused respectively, and the average value of the propagation time, the fused amplitude, the fused speed and the fused coordinates corresponding to each transducer are used to form the attribute of the detection point; mapping each detection point attribute to the three-dimensional coordinate system of the oil tank, and obtaining the detection sparse network based on the graph structure sparse method.

[0036] In the present application, a plurality of transducer arrays are arranged on the surface of the transformer oil tank in advance, the transducer arrays are used to emit continuous ultrasonic pulse signals to different directions inside the oil tank, and transmit and receive the transmitted waves and reflected waves propagated through the medium inside the oil tank. The ultrasonic transducer suitable for detecting the temperature of the transformer oil is selected; according to the actual appearance data and internal structure of the transformer, the ultrasonic transducer is arranged reasonably around the transformer oil tank, and the transmitting and receiving transducers are placed at the same height of the shell and correspond to each other; in the embodiment, a plurality of groups of ultrasonic transducers are arranged around the transformer oil tank, each group of ultrasonic transducers includes one transmitting transducer and one receiving transducer; the master control module controls the operation of each group of ultrasonic transducers to realize the propagation of ultrasonic waves in different propagation directions in the transformer oil tank; coupling agent is applied on the surface of the ultrasonic transducer to ensure close contact with the shell, so that the ultrasonic wave can penetrate the shell and the insulating oil and be received by the receiving transducer which is also coated with coupling agent; in the embodiment, the ultrasonic transducer suitable for detecting the transformer oil tank is selected, and according to the actual appearance data and internal structure of the transformer, the ultrasonic transducer is arranged reasonably around the transformer oil tank, and the transmitting and receiving transducers are placed at the same height of the shell and correspond to each other. Coupling agent is applied on the surface of the ultrasonic transducer to ensure close contact with the shell, so that the ultrasonic wave can penetrate the shell and the insulating oil and be received by the receiving transducer which is also coated with coupling agent.

[0037] As a detection means, ultrasonic waves are used to realize the conversion between electrical signals and ultrasonic signals, and a device for generating and receiving ultrasonic waves is required. The ultrasonic probe is one of the core components of the hardware part of the detection device. Reasonable selection of the size, transmission frequency, piezoelectric material and other parameters of the ultrasonic probe can greatly improve the measurement accuracy. From the structure and electromechanical energy conversion form of the transducer, the ultrasonic probe should have the following characteristics or functions: compact structure, easy to install; the frequency is suitable for propagation in solids and liquids; the working frequency is designed at a lower frequency band to reduce the energy absorption and attenuation during sound wave propagation; and a larger radiation surface is provided to make the transducer have good directivity.

[0038] According to the required penetration distance of the to-be-measured, the center frequency of the ultrasonic transducer selected in the embodiment is 400KHz, which can meet the actual detection requirements. On the basis of ensuring that the ultrasonic transducer can penetrate the structure of the transformer shell, the ultrasonic transducer has a smaller diffusion angle, so that the ultrasonic energy is concentrated and the directivity is better, which provides strong hardware equipment support for subsequent signal algorithm processing.

[0039] The master control module controls the emission transducer to emit ultrasonic waves and controls the receiving transducer to receive ultrasonic waves; when the emission transducer emits ultrasonic waves, the timer inside the master control module starts timing; when the receiving transducer receives ultrasonic waves, the timer inside the master control module stops timing; the data of the timer is the propagation time of ultrasonic waves on different propagation paths. When the emission transducer emits ultrasonic waves, the timer inside the master control module starts timing; when the receiving transducer receives ultrasonic waves, the timer inside the master control module stops timing; the data of the timer is the propagation time of ultrasonic waves on different propagation paths.

[0040] The emission and reception of multiple groups of ultrasonic transducers are determined by the emission-reception switching control circuit of the master control module, so as to control the propagation direction of ultrasonic waves. In the embodiment, the master control module includes an FPGA; the FPGA emits a 3.3V 400kHz square wave pulse signal, the square wave pulse signal is converted into a high-voltage pulse signal of several hundred volts by a high-voltage driving circuit, the high-voltage pulse signal excites the emission transducer to emit 400kHz ultrasonic waves, the internal timer of the FPGA starts timing when the emission transducer emits ultrasonic waves, and stops timing when the receiving transducer receives ultrasonic waves; the data of the timer is the propagation time of ultrasonic waves in the transformer oil tank.

[0041] The control signal emitted by the master chip controls the on-off of MOS tube Q1, so that the voltage applied to the primary side of the transformer changes. The role of MOS tube is critical, so it should be selected as a high-speed MOS switch tube to avoid distortion of the excitation signal; at the same time, it also plays a role in isolating and protecting the master chip, which can effectively avoid the large current of the primary side of the transformer from flowing back to the I / O port of the master chip to damage the master chip. When the voltage of the primary coil changes with the excitation signal, the magnetic flux of the coil changes, causing the secondary coil to induce a voltage. Select the matching transformer, whose working frequency matches the transducer. The low-voltage power supply provides input power for the transformer, so the output voltage can reach a higher peak-to-peak value. In this way, the power of the excitation signal is amplified, which can provide the required AC voltage and power for the transducer. If you want to drive the ultrasonic transducer to emit ultrasonic waves, in addition to the AC voltage, a large DC bias is also needed. Here, the energy storage of capacitor C1 and the clamping effect of diode D1 are used to provide a higher DC voltage. To prevent C1 and D1 from being broken down, a high-voltage capacitor and diode should be selected, and the capacitance of C1 should be much larger than the equivalent capacitance of the transducer, so that the AC excitation voltage can mainly fall on the transducer. Finally, the AC excitation voltage and the DC bias are superimposed to form a higher driving voltage to drive the transducer to work. In the embodiment, the echo is processed by the filter circuit and the amplification circuit of the signal processing module, and is uploaded to the host computer through the serial communication for data processing; according to the time of flight corresponding to each flight path collected by the hardware system; the echo signal conditioning circuit mainly includes a filter module, a self-gain module and a hysteresis comparison level conversion module. The ultrasonic echo signal is converted into a digital signal that meets the level standard of the master chip for subsequent processing and calculation of the master chip.

[0042] The ultrasonic echo signal inevitably contains some noise, so before processing the echo, these useless noises should be filtered out to avoid affecting the quality of the echo signal. Therefore, according to the frequency band characteristics of the ultrasonic transducer, a band-pass filter is designed for it, which requires that the response in the passband should be as flat as possible, and the upper and lower boundary attenuation should be as steep as possible.

[0043] The ultrasonic wave attenuates exponentially in the medium, and the longer the propagation distance, the greater the attenuation, so the amplitude of the ultrasonic wave reflected from different distances varies greatly. Since the project uses a correlation algorithm to capture the time of flight, the closer the shape of the echo to the standard echo stored in advance, the more accurate the calculation result. Therefore, an automatic gain circuit (AGC) should be designed, which can adaptively adjust the amplification factor according to the amplitude of the echo, so that the amplitude of the echo from different distances can be adjusted to the same as the reference signal and meet the level standard of the subsequent processing circuit. AGC is a key part of the signal conditioning circuit system.

[0044] The polarization circuit design employs a hysteresis comparator circuit to polarize the self-gained analog echo signal into a binary signal, conforming to the I / O port level standards of the main control chip. Compared to zero-crossing comparators, the hysteresis comparator can filter out noise and ripple interference while binarizing the signal. The threshold voltage can be adjusted using a sliding rheostat, allowing for the setting of a reasonable threshold value based on the noise conditions in the actual environment.

[0045] By performing short-time Fourier transform time-frequency analysis and waveform feature processing on the ultrasonic echo signal, noise interference is eliminated and signal stability is enhanced, thereby enabling the extraction of stable sound velocity estimates and spatial coordinate information from multipath data. The received echo is subjected to short-time Fourier transform, and an energy mask is applied in the time-frequency domain before inverse transform to obtain a denoised signal.

[0046] By employing a multi-path coverage arrangement, ultrasonic signals can be acquired from multiple spatial regions within the fuel tank, obtaining parameters such as propagation time, amplitude variations, and waveform characteristics under different paths. This forms a sparsely distributed set of sound velocity observation points, including: transducer Reflection point and transducer A path , , The number of transducers, , For transducers The number of corresponding reflection points; transducer Received path The amplitude of the ultrasonic signal is shown in the following formula:

[0047] In the formula, , Paths The amplitude and propagation time of the ultrasonic signal on the surface. For transducers The initial amplitude of the emitted ultrasonic signal, To comprehensively assess the attenuation factor, which reflects the absorption and scattering effects of the medium, the value in the examples is taken as 0.78~0.92; In the embodiments, it is a non-limiting but preferred choice that the amplitude of the ultrasonic wave decays exponentially with the propagation time when it propagates in the oil tank medium. When the reflection time of the ultrasonic echo signal received by the transducer is known, the reflection or transmission amplitude of the ultrasonic wave at different oil temperatures of the transformer can be inferred from this.

[0048] Based on transducer The propagation time of all received ultrasonic signals is fused with the amplitude of all ultrasonic signals, as shown in the following formula:

[0049] wherein, is the transducer is the fused amplitude of all the received ultrasonic signals; The geometric length of each path is determined by geometric method according to the emitting angle of each transducer and the size of the transformer oil tank; The speed of the ultrasonic signal on the corresponding path is shown as follows:

[0050] wherein, , is the speed of the ultrasonic signal on the path and the geometric length of the path, is the denoised ultrasonic time-domain signal, is the transmitting pulse of the transducer, is the transmitting period; The propagation time of all the received ultrasonic signals based on the transducer is fused to the transducer The speed of all the received ultrasonic signals based on the transducer

[0051] wherein, is the fused speed of all the received ultrasonic signals based on the transducer ; The geometric barycentric coordinates of all the path intersection segments of the transducer are determined by geometric barycentric method; wherein, is the path intersection point; The propagation time of all the received ultrasonic signals based on the transducer is fused to the geometric barycentric coordinates of all the path intersection segments of the transducer ;

[0052] wherein, is the fused coordinates of the geometric barycenter of all the path intersection segments of the transducer ; The corresponding propagation time average , fused amplitude , fused speed , fused coordinates of the transducer constitute the detection point attribute , and the detection point set is The attribute of each detection point is mapped into the three-dimensional coordinate system of the oil tank, and a detection sparse network is obtained based on a graph structure sparse method.

[0053] The propagation time equivalent on different paths is taken as a weighting coefficient based on path effect, and the multi-item data fusion of the ultrasonic wave signals on different paths is weighted and processed, the consistency of the paths is evaluated through the fusion weighted variance of the data, and a sparse distribution of the sound velocity observation point set is formed; in addition, stable sound velocity estimation values and spatial coordinate information are extracted from the multi-path data, and the spatial coordinates and the sound velocity estimation values of the observation points are mapped into the three-dimensional coordinate system of the oil tank as node attributes to form a sparse graph structure point set network with geometric topological relationship and physical measurement information, which is used for subsequent interpolation and reconstruction of the temperature field.

[0054] In step 2, the oil tank is divided into multiple isometric layers along the height direction of the oil tank, and a thin plate spline base function of each layer is established; based on the thin plate spline base function of each layer, a thin plate spline interpolation fitting is performed on the detection sparse network with the minimum bending energy of the thin plate as an optimization target, and a global detection network is obtained; the speed estimation value of the ultrasonic wave signal corresponding to any point in the oil tank is determined by using the high confidence representative value and the low confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network and the radial basis function of the thin plate spline base, so as to establish the sound velocity distribution field in the oil tank; according to the sound velocity distribution field in the oil tank, the corrected oil temperature estimation value is obtained by using the Bayesian statistical method based on the oil temperature observation model of Gaussian likelihood.

[0055] The oil tank is divided into multiple isometric layers along the height direction of the oil tank, and the height of each layer is not greater than the diameter of the effective coverage area of the signal emitted by a single transducer. The center point coordinates of the i-th layer are , , The number of layers is n; the distance between any point in the i-th layer and the center point is . The geometric center point of the i-th layer is taken as a reference to construct the thin plate spline base function of the i-th layer . The thin plate spline base function of the i-th layer is constructed as follows:

[0056] The integral values of the thin plate spline base functions of each layer on different paths are used to construct a propagation matrix , The number of transducers is m, and the number of reflection points corresponding to one transducer is n. Based on the propagation matrix, the fusion speed corresponding to the transducer is used to establish a sound velocity observation vector as follows: ​​​

[0057] wherein, is a position coordinate, is a transducer is a sound speed observation vector, is a transducer corresponding fusion speed; a thin plate bending energy is established based on the sound speed observation vector, as shown in the following formula:

[0058] wherein, is a thin plate bending energy, is a second-order multi-index, is a second-order partial derivative; a thin plate spline basis function under corresponding fusion coordinate constraints is established based on the transducer a global detection network is obtained by thin plate spline difference fitting of the detection sparse network, and a detection point attribute obtained by the thin plate spline difference fitting is determined , , is a number of detection points obtained by the thin plate spline difference fitting.

[0059] The present application constructs a thin plate spline basis function interpolation model by taking the extracted sparse sound speed points as constraints, and generates a smooth and continuous sound speed field by globally fitting the sparse points in a three-dimensional space by taking the minimum bending energy of the thin plate spline basis function as an optimization criterion.

[0060] The signal-to-noise ratio is calculated by using the fusion amplitude of each detection point in the global detection network, and the detection points with a signal-to-noise ratio not less than a set threshold are high-confidence points, and the detection points with a signal-to-noise ratio less than a set threshold are low-confidence points; a high-confidence interval is , and a low-confidence interval is , is a number of high-confidence points; a median of the high-confidence interval is taken as a high-confidence representative value , and a median of the low-confidence interval is taken as a low-confidence representative value ; Each detection point in the global detection network is sorted in order from small to large according to the signal-to-noise ratio, and the ratio of the sorting serial number of each detection point to + is taken as the confidence weight of each detection point, so as to establish a confidence weight diagonal matrix ; A weighted objective function of the high-confidence representative value and the low-confidence representative value is established, as shown in the following formula:

[0061] In the formula, Let be the weighted objective function for high-confidence representative values ​​and low-confidence representative values. This represents the sound speed observation vector with high confidence. The sound speed observation vector represents the low-confidence value. Let be the vector formed by the fusion velocities of all transducers. For regularization terms; With the goal of minimizing the change in the weighted objective function, high-confidence representative values ​​and low-confidence representative values ​​are determined. This invention uses weighting factors to differentiate between high-confidence and low-confidence points, achieving adaptive error suppression. This can suppress measurement errors and thus obtain more stable and reliable three-dimensional sound velocity field reconstruction results.

[0062] The estimated vector of the speed of sound at any point inside the fuel tank is shown in the following equation:

[0063] In the formula, Let be the estimated vector of the sound speed at any point in the fuel tank. Let be the coordinate vector of any point in the fuel tank. For the fuel tank The coordinate vector of the centroid of the layer is fitted. For constant bias terms, For the radial basis functions of the thin plate spline basis, , Let any point in the fuel tank be connected to the fuel tank's first... The distance between the centroids of the fitted layer. The value is 0.001; Using the propagation matrix and the sound velocity estimation vector at any point inside the tank, the velocity estimate of the ultrasonic signal at any point inside the tank is obtained, as shown in the following formula:

[0064] In the formula, This is the estimated velocity of the ultrasonic signal at any point inside the fuel tank. Based on the velocity estimate of the ultrasonic signal corresponding to any point inside the fuel tank, the sound velocity distribution field of the fuel tank is established. The initial value of the predicted temperature is obtained by mapping the linear multivariate physical relationship between speed and oil temperature. Establish the Bayesian prior distribution of oil temperature. , , Based on The sound velocity fusion kernel matrix of thin plate spline basis. For smoothing coefficients, The prior covariance matrix; Based on the oil temperature Bayesian prior distribution, the thin plate bending energy is ; An oil temperature observation model based on Gaussian likelihood is established, as shown in the following formula: ,

[0065] In the formula, is the observed temperature measurement data, is the real temperature field of the oil tank, is the measurement noise, is the noise covariance matrix; The corrected oil temperature estimation value is obtained by closed-form Bayesian updating of the posterior mean value, as follows:

[0066] In the formula, is the corrected oil temperature estimation value.

[0067] The sound speed field obtained by fitting the thin plate spline basis function is converted into a temperature distribution, and the mapping is completed according to the linear multivariate physical relationship model of sound speed-temperature. Finally, a continuous three-dimensional oil temperature temperature field is formed in the three-dimensional space, and is further optimized through Bayesian inference correction, so that the reconstruction result not only conforms to the observation data, but also maintains physical rationality and spatial smoothness.

[0068] The application also provides a layered detection system for transformer oil temperature based on a thin plate spline basis function, comprising: A detection data processing module is configured to obtain the propagation time, amplitude and speed of ultrasonic signals on different paths corresponding to each transducer, and obtain the coordinates of the geometric centers of all path intersection segments corresponding to each transducer; fuse the amplitude, speed and coordinates based on the propagation time, respectively, to form detection point attributes including the average propagation time, fused amplitude, fused speed and fused coordinates corresponding to each transducer; map the detection point attributes to an oil tank three-dimensional coordinate system, and obtain a detection sparse network based on a graph structure sparse method; An oil temperature estimation module is configured to divide the oil tank into multiple isometric layers along the height direction of the oil tank, and establish thin plate spline basis functions for the layers; based on the thin plate spline basis functions for the layers, perform thin plate spline interpolation fitting on the detection sparse network to obtain a global detection network, by taking the minimum bending energy of the thin plate as an optimization target; determine the speed estimation value of the ultrasonic signal corresponding to any point in the oil tank by using the high confidence representative value and the low confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network, and the radial basis function of the thin plate spline basis, to establish a sound speed distribution field of the oil tank; and obtain a corrected oil temperature estimation value based on a Gaussian likelihood-based oil temperature observation model and a Bayesian statistical method according to the sound speed distribution field of the oil tank.

[0069] In the prior art, a least square temperature field reconstruction algorithm uses a least square method to obtain temperature values of a limited number of points and reconstructs a temperature field of the entire to-be-measured region through interpolation. However, this method has a problem that temperature information at a boundary of the to-be-measured region cannot be obtained because of a limitation in interpolation. As a result, the temperature information at the boundary of the to-be-measured region is lost in the reconstruction result, and the temperature distribution characteristics of the entire to-be-measured region cannot be accurately reflected.

[0070] In the embodiment, the obtained main graph of the three-dimensional temperature field of the transformer oil tank is as shown in Figure 2 , the rotation graph of the three-dimensional temperature field of the transformer is as shown in Figure 3 , and the overhead view of the temperature field of the transformer is as shown in Figure 4 . The present application pre-arranges a plurality of transducer arrays on the surface of the transformer oil tank, collects ultrasonic signals of a plurality of spatial regions inside the oil tank to form a sparse distribution of a set of sound velocity observation points, processes the ultrasonic echo signals through time-frequency analysis and waveform feature processing to form a network of sparse graph structure points with geometric topological relationship and physical measurement information, which is used for subsequent difference reconstruction of the temperature field. An oil temperature distribution model of the height of the oil tank is established according to the echo physical information, and each layer is evenly divided into a plurality of isometric sub-layers according to the diameter of the effective coverage area of a single ultrasonic transducer. With the extracted sparse sound velocity points as constraints, a thin plate spline basis function interpolation model is constructed to globally fit the sparse points in the three-dimensional space, solving the problem of the actual collection points caused by the limited measurement in the space; at the same time, the high confidence and low confidence points are differentially processed by using a weighting factor, so that a more stable and reliable three-dimensional sound velocity field reconstruction result is obtained. The sound velocity field obtained through the thin plate spline basis function fitting is converted into a temperature distribution, and the mapping is completed according to a linear multivariate physical relationship model of sound velocity-temperature, realizing the effect of three-dimensional oil temperature field reconstruction by ultrasonic measurement. The problem of temperature information loss in the least square temperature field reconstruction algorithm is effectively overcome, thereby improving the overall performance of the temperature field reconstruction algorithm.

[0071] Based on the temperature field distribution function of the transformer oil tank, the layered detection result of the transformer oil temperature further includes: 1) The temperature field of the transformer oil tank is visually displayed through the MATLAB GUI software.

[0072] 2) The temperature distribution of each measurement layer inside the transformer oil tank, the temperature of the overheated part, and the position coordinates of the overheated part are displayed through different perspectives, specifically as follows: ① Three-dimensional temperature field overall graph: the system generates a three-dimensional temperature field of the transformer oil tank according to the imported coordinates and corresponding temperature data inside the transformer oil tank, and displays the overall distribution graph of the temperature field. Since the three-dimensional temperature field result is four-dimensional information, in order to facilitate display, the system presents the temperature condition of the temperature field through a slice graph.

[0073] ②Three-dimensional temperature field layered diagram: the system divides the three-dimensional temperature field of the transformer oil tank into four characteristic layers of top layer, middle layer, bottom layer and highest temperature layer according to the imported position and temperature data, which is beneficial for the user to intuitively and clearly observe the temperature distribution in the oil tank.

[0074] ③Temperature field flat diagram: in order to display the temperature field from multiple angles, the system displays the distribution of the three-dimensional temperature field of the transformer from the overhead perspective.

[0075] ④Temperature field grid diagram: in order to display the temperature field from multiple angles, the system displays the distribution of the three-dimensional temperature field of the transformer through the contour map.

[0076] The present disclosure can be a system, a method, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0077] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched-tape, a holographic storage medium, or any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0078] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0079] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0080] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, but not to limit it. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A hierarchical detection method of transformer oil temperature based on thin plate spline basis function, a plurality of transducers are arranged on the surface of the oil tank, each transducer receives ultrasonic wave signals in different paths; characterized in that, The method comprises: acquiring the propagation time, amplitude and speed of the ultrasonic signal on different paths corresponding to each transducer, and acquiring the coordinates of the geometric center of intersection of all paths corresponding to each transducer; fusing the amplitude, speed and coordinates based on the propagation time to form the detection point attribute of the average propagation time, fused amplitude, fused speed and fused coordinates corresponding to each transducer; mapping the detection point attribute to the three-dimensional coordinate system of the oil tank, and obtaining a detection sparse network based on a graph structure sparse method; dividing the oil tank into multiple isohypse layers along the height direction of the oil tank, and establishing a thin plate spline basis function of each layer; based on the thin plate spline basis function of each layer, a thin plate spline interpolation fitting is performed on the detection sparse network to obtain a global detection network, by taking the minimum bending energy of the thin plate as an optimization target; the speed estimation value of the ultrasonic signal corresponding to any point in the oil tank is determined by using the high confidence representative value and the low confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network, and the radial basis function of the thin plate spline basis, so as to establish the sound speed distribution field in the oil tank; and the corrected oil temperature estimation value is obtained by using the Bayesian statistical method based on the oil temperature observation model of the Gaussian likelihood.

2. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 1, characterized in that transducer , reflection points and transducers constitute a path , , the number of transducers, , transducers the number of corresponding reflection points; transducers paths the amplitude of the ultrasonic signal received is given by wherein , are the amplitude and the propagation time of the ultrasonic signal on the path , is the initial amplitude of the ultrasonic signal emitted by the transducer , is the overall attenuation factor; Transducer-based The propagation times of all received ultrasound signals are fused with the amplitudes of all ultrasound signals as shown in the following equation: wherein transducer fusion amplitude of all received ultrasound signals.

3. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 2, characterized in that The geometric length of each path is determined using geometric methods based on the angle of emission of each transducer and the dimensions of the transformer tank; the transducers The speed of the ultrasonic signal on each path is given by the following equation: wherein , are the velocity of the ultrasonic signal on the path and the geometrical length of the path, respectively, is the denoised ultrasonic time domain signal, is the transmit pulse of the transducer, is the transmit period; Based on the transducer The propagation times of all received ultrasound signals are fused The velocities of all received ultrasound signals are fused as follows: wherein transducer fusion velocity of all received ultrasound signals.

4. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 3, characterized in that The transducer is determined using the geometric centroid method. Geometric centroid coordinates of all path intersections ,in, For path intersections; based on transducers The propagation time of all received ultrasonic signals to the transducer The geometric centroid coordinates of all path intersections are merged as shown in the following formula: wherein is the fused coordinate of the geometric center of all path intersection segments of the transducer .

5. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 1, characterized in that Transducer Corresponding propagation time average , fusion amplitude , fusion speed , fusion coordinates , constitute the detection point attribute , detection point set is ; the detection point attribute is mapped to the three-dimensional coordinate system of the oil tank, and a detection sparse network is obtained based on a graph structure sparse method; Transducer-based Corresponding to the fusion coordinate constraint under the thin plate spline basis function, the thin plate spline difference fitting is carried out on the detection sparse network, the global detection network is obtained, and the detection point attribute obtained by the thin plate spline difference fitting is determined , , The number of detection points obtained by the thin plate spline difference fitting.

6. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 5, characterized in that Along the height of the fuel tank, the tank is divided into multiple layers of equal height, with the height of each layer not exceeding the diameter of the effective coverage area of ​​a single transducer's transmitted signal; The coordinates of the center point of the layer are , , The number of layers; any point within a layer The distance from the center point is , with the first Based on the geometric center point of the layer, construct the first... Thin-plate spline basis functions As shown in the following formula: The propagation matrix is constructed by the integral values of the thin-plate spline basis functions of each layer on different paths , is the number of transducers, is the number of reflection points corresponding to one transducer.

7. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 6, characterized in that Based on the propagation matrix, the transducer A sound speed observation vector is established corresponding to the fusion velocity, as shown in the following formula: wherein is a position coordinate of a transducer is a sound speed observation vector of a transducer is a corresponding fused velocity; the bending energy of the thin plate is established based on the sound speed observation vector, and is shown in the following formula: wherein is the sheet bending energy, is the second order multiplicative index, is the second order partial derivative.

8. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 7, characterized in that the signal-to-noise ratio is calculated by using the fused amplitude of each detection point in the global detection network, the detection points with the signal-to-noise ratio not less than a set threshold are high confidence points, and the detection points with the signal-to-noise ratio less than the set threshold are low confidence points; High confidence interval is , low confidence interval is , Number of high confidence points; median of high confidence interval as high confidence representative value Median of low confidence interval as low confidence representative value ; The detection points in the global detection network are sorted in order of signal-to-noise ratio from small to large, and the ratio of the sorted serial number of each detection point to the total number of detection points is taken as the confidence weight of each detection point, so as to establish a confidence weight diagonal matrix + ;​ a weighted objective function of the high confidence representative value and the low confidence representative value is established, and is shown in the following formula: wherein is a weighted objective function of high-confidence representative values and low-confidence representative values, is a sound speed observation vector of high-confidence representative values, is a sound speed observation vector of low-confidence representative values, is a vector of fused speed of all transducers, is a regularization term; the high confidence representative value and the low confidence representative value are determined by taking the minimum change of the weighted objective function as an optimization target.

9. The layered detection method for the transformer oil temperature based on the thin plate spline basis function according to claim 8, characterized in that the sound speed estimation vector of any point in the oil tank is shown in the following formula: wherein, is the sound speed estimation vector at any point in the tank, is the coordinate vector at any point in the tank, is the coordinate vector of the fitted barycenter point of the first layer of the tank, is the constant bias term, is the radial basis function of the thin plate spline basis, , is the distance between any point in the tank and the fitted barycenter point of the first layer of the tank, is set to 0.001; the speed estimation value of the ultrasonic signal corresponding to any point in the oil tank is obtained by using the propagation matrix and the sound speed estimation vector of any point in the oil tank, and is shown in the following formula: In the formula, is the velocity estimate of the ultrasonic signal corresponding to any point in the oil tank; Based on the speed estimation value of the ultrasonic signal corresponding to any point in the tank, the sound speed distribution field of the tank is established.

10. The transformer oil temperature layered detection method based on the thin plate spline basis function according to claim 9, characterized in that, Mapping based on linear multivariate physical relationship between velocity and oil temperature to get the initial value of predicted temperature ; establishing the Bayesian prior distribution of oil temperature , , for the velocity fusion velocity kernel matrix based on the thin plate spline basis of , is the smoothing coefficient, is the prior covariance matrix; Based on the Bayesian prior distribution of oil temperature, the thin-plate bending energy is ; An oil temperature observation model based on Gaussian likelihood is established, as shown in the following formula: ， wherein is the observed temperature measurement data, is the real temperature field of the tank, is the measurement noise, is the noise covariance matrix; The corrected oil temperature estimation value is obtained by closed-form Bayesian updating of the posterior mean value, as shown in the following relationship: In the formula, is the corrected oil temperature estimation value.

11. A hierarchical detection system of transformer oil temperature based on thin plate spline basis function, for implementing the hierarchical detection method of transformer oil temperature based on thin plate spline basis function according to any one of claims 1 to 10; characterized in that, Comprise: The detection data processing module is configured to obtain the propagation time, amplitude and speed of the ultrasonic signal on different paths corresponding to each transducer, and obtain the coordinates of the geometric center of intersection of all paths corresponding to each transducer; based on the propagation time, the amplitude, speed and coordinates are fused respectively, and the propagation time average value, fused amplitude, fused speed and fused coordinates corresponding to each transducer constitute the detection point attribute; The detection point attribute is mapped to the three-dimensional coordinate system of the tank, and a detection sparse network is obtained based on the graph structure sparse method; The oil temperature estimation module is configured to divide the tank into multiple isohypse layers along the height direction of the tank, and establish a thin plate spline basis function for each layer; based on the thin plate spline basis function of each layer, the detection sparse network is thin plate spline interpolation fitted with the minimum bending energy of the thin plate as the optimization target, to obtain a global detection network; the speed estimation value of the ultrasonic signal corresponding to any point in the tank is determined by using the high confidence representative value and the low confidence representative value determined based on the signal-to-noise ratio of each detection point in the global detection network, and the radial basis function of the thin plate spline basis, to establish the sound speed distribution field of the tank; based on the Gaussian likelihood of the oil temperature observation model, the corrected oil temperature estimation value is obtained by using the Bayesian statistical method according to the sound speed distribution field of the tank.

12. A terminal comprising a processor and a storage medium; characterized in that: The storage medium is configured to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method of any one of claims 1-10.

13. A computer readable storage medium having stored thereon a computer program, characterized in that The program is executed by the processor to implement the steps of the method of any one of claims 1-10.

Images