Winding hot spot temperature nondestructive detection method and system based on active ultrasonic emission
By establishing an ultrasonic propagation model and a reverse propagation neural network in the transformer, the problem of accurately detecting the hot spot temperature inside the transformer was solved, and high-precision non-destructive detection of winding hot spot temperature was achieved.
Patent Information
- Application Number
- CN202310516240.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-09
- Publication Date
- 2025-12-09
- Estimated Expiration
- 2043-05-09
AI Technical Summary
Existing methods for measuring the internal temperature of transformers have problems such as limited measurement range, low accuracy, and difficulty in long-term effectiveness. In particular, in transformers with complex structures, the propagation path of ultrasonic waves is difficult to identify, making it difficult to detect hot spot temperatures.
An ultrasonic propagation model was established. By setting ultrasonic sensors in the transformer geometric model, active ultrasonic emission was performed. Combined with a backpropagation neural network, the characteristic parameters of the acoustic signal were extracted, and the backpropagation neural network was trained to achieve non-destructive detection of the winding hot spot temperature.
This improved the accuracy of detecting hot spot temperatures in transformer windings, reduced the impact of regional information on temperature calibration accuracy, and achieved higher accuracy in temperature diagnosis.
Smart Images
Figure CN116858396B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of winding hot spot temperature detection, in particular to a winding hot spot temperature nondestructive detection method and system based on active ultrasonic emission. BACKGROUND
[0002] As an important device for power transmission and power distribution, the safe and stable operation of power transformers plays an important role in the operation of power systems. Although oil-immersed transformers are widely used due to their relatively good heat dissipation performance, low loss, large capacity and other advantages, the internal loss and heating phenomenon of the transformer still need to be focused on during operation. Studies have shown that the operating condition of the transformer is closely related to the change of the internal temperature field of the transformer, and the winding hot spot temperature of the transformer, as an important parameter reflecting the state of the transformer, must be controlled within a reasonable range. High temperature will increase the aging rate of the transformer and even damage the insulation level of the transformer. Therefore, the detection of the internal hot spot temperature of the transformer can be used as an effective method for transformer accident prevention and fault diagnosis.
[0003] The current transformer internal temperature measurement methods include direct temperature measurement method and indirect temperature measurement method. In the direct temperature measurement method, the oil surface thermometer temperature measurement method and the infrared temperature measurement method directly measure the top layer oil temperature and the shell temperature of the transformer, respectively, and the measurement range is very limited. The thermal resistance temperature measurement and the thermocouple temperature measurement are difficult to guarantee long-term effective measurement accuracy due to the corrosion of metal wires and other problems. The optical fiber temperature sensor measurement method is difficult to apply in the transformer that has been put into operation due to the problem of embedding sensors in the winding.
[0004] As a method for indirectly measuring the temperature of a medium by using the functional relationship between the temperature of the medium and the propagation characteristics of the acoustic wave, ultrasonic temperature measurement has the advantages of high precision, real-time continuous measurement, and convenient maintenance. And in the field of transformers, although there is no relevant research on ultrasonic temperature measurement technology, ultrasonic waves as a non-invasive detection technology have been applied in the field of transformer internal detection due to their advantages such as not being affected by electromagnetic interference and high safety. However, unlike the application of ultrasonic waves in other temperature measurement fields, the internal structure of the transformer is complex and does not meet the conditions of a single medium. The problems of multipath transmission and multiple reflection between media exhibited by the sound wave in the propagation process make it difficult to identify the sound wave propagation path. Therefore, it is urgent to develop an effective detection method that can obtain the internal hot spot temperature of the transformer without affecting the internal structure. SUMMARY
[0005] To solve the above problems, the first aspect of the present application provides a winding hot spot temperature nondestructive detection method based on active ultrasonic emission, which specifically includes the following steps:
[0006] S1, establishing an ultrasonic wave propagation model based on the propagation process of sound waves between different media;
[0007] S2, establishing a geometric model of the transformer, and setting a plurality of ultrasonic sensors on the surface of the box of the geometric model, determining the sound source signal and excitation voltage of the ultrasonic sensors;
[0008] S3, simulating hot spot faults by setting point heat sources in the winding of the geometric model, simulating the temperature field of the hot spot faults at different positions at different temperatures, and obtaining a temperature field distribution data set;
[0009] S4, based on the internal temperature field distribution of the geometric model of the transformer, using the ultrasonic wave propagation model to actively emit ultrasonic waves, obtaining the sound pressure distribution diagram of the ultrasonic waves passing through the temperature field, and obtaining a data set of the sound pressure distribution;
[0010] S5, based on the sound field simulation of the ultrasonic wave propagation model in the temperature field, using the ultrasonic detection method of one emission and multiple receptions to obtain the sound wave signal, and extracting the effective signal of the sound wave according to the data set of the sound pressure distribution;
[0011] S6, extracting signal characteristic parameters according to the effective signal of the sound wave, training a back propagation neural network, establishing an inversion database of hot spot area-hot spot temperature, and detecting and classifying the hot spot of the winding to be measured through the trained back propagation neural network, and obtaining the winding hot spot information.
[0012] In some implementations of the first aspect, the signal characteristic parameters are extracted from the effective signal of the sound wave, and the types of the signal characteristic parameters include peak time, time domain index and frequency domain index.
[0013] In some implementations of the first aspect, the back propagation neural network is trained, and the specific steps include:
[0014] establishing temperature field samples of different hot spot areas and different heat source power cross combinations, the temperature field samples including signal characteristic parameters of a plurality of effective signals of sound waves;
[0015] establishing a region diagnosis model and a temperature diagnosis model of the back propagation neural network, and fitting and training the mapping relationship between the temperature field samples of the hot spot faults and the signal characteristic parameters;
[0016] The region diagnosis model is used to classify the hot spot fault region first, and then the temperature diagnosis model corresponding to a single region is used to classify the heat source power.
[0017] In some implementations of the first aspect, the sound source signal of the ultrasonic sensor is determined, and the specific steps include that the sound source signal is a continuous wave of 10 cycles; and the amplitude of the excitation voltage is 200V.
[0018] Further, the data set of the sound pressure distribution is extracted from the sound wave effective signal, and a specific method is that the sound wave effective signal is after the sound wave starting signal, the starting time and the ending time of the sound wave effective signal have amplitude mutations, the starting time is the time when the rising amplitude of the trough-peak is the largest, the ending time is the time when the falling amplitude of the peak-trough is the largest, and 10 continuous waves are included between the starting time and the ending time.
[0019] In some implementations of the first aspect, the hot spot fault is simulated by setting a point heat source in the geometric model, and a specific method is to set x hot spot fault positions and y hot spot fault temperatures, and a total of x*y types of hot spot fault types are set.
[0020] In some implementations of the first aspect, the ultrasonic wave propagation model is established based on the propagation process of the sound wave between different media inside the transformer, and specifically includes setting the sound pressure fluctuation equation in the fluid and the solid and the sound-solid coupling boundary condition to establish the ultrasonic wave propagation model.
[0021] The second aspect provides a winding hot spot temperature non-destructive detection system based on active ultrasonic emission, comprising:
[0022] The ultrasonic wave propagation model establishment module is configured to establish an ultrasonic wave propagation model based on the propagation process of the sound wave between different media inside the transformer.
[0023] The transformer model module is configured to establish a geometric model of the transformer, and set a plurality of ultrasonic wave sensors on the surface of the box of the geometric model, and determine the sound source signal and the excitation voltage of the ultrasonic wave sensors.
[0024] The temperature field distribution module is configured to simulate the temperature field of the hot spot fault at different positions and at different temperatures by setting a point heat source in the geometric model, and obtain a temperature field distribution data set.
[0025] The sound pressure distribution module is configured to perform active ultrasonic emission by using the ultrasonic wave propagation model based on the temperature field distribution inside the geometric model of the transformer, obtain a sound pressure distribution diagram of the ultrasonic wave passing through the temperature field, and obtain a data set of the sound pressure distribution.
[0026] The sound wave signal acquisition module is configured to simulate the sound field propagation of the temperature field based on the ultrasonic wave propagation model, acquire the sound wave signal by using the ultrasonic detection method of one emission and multiple receptions, and extract the sound wave effective signal according to the data set of the sound pressure distribution.
[0027] The inversion processing module is configured to extract signal characteristic parameters according to the sound wave effective signal, train the back propagation neural network, establish an inversion database of the hot spot area-hot spot temperature, perform detection and classification on the to-be-measured winding hot spot by using the trained back propagation neural network, and obtain the winding hot spot information.
[0028] Further, the transformer model module comprises a steel box, a transformer winding and transformer oil.
[0029] The hot spot fault positions are uniformly arranged at the same height position in the upper part of the transformer winding and are located outside the transformer winding.
[0030] The beneficial effects are:
[0031] (1) Based on the finite element principle, a sound field simulation model of the transformer is established. A point heat source is added in the model to simulate the winding hot spot, and the temperature field distribution corresponding to the hot spot fault is obtained. On the basis of a certain hot spot fault temperature field, by setting the sound pressure fluctuation equation in fluid and solid and the sound-solid coupling boundary condition, the sound wave propagation process in the transformer is simulated, and the ultrasonic signals received by the sensor are recorded, which are used to construct the data set corresponding to the hot spot.
[0032] (2) By changing the heat source power and the heat source position, the steady-state temperature field distribution of the hot spot under the cross combination of x fault positions and y heat source powers is obtained, a total of x*y temperature field samples. According to the influence law of the temperature field on the sound wave propagation, the characteristic parameters of the sound wave effective signal are extracted. The reverse propagation neural network is designed to realize the preliminary inversion of the hot spot fault, and finally the detection method of the hot spot area-hot spot temperature is determined. Through the detection method of determining the hot spot area first and then determining the hot spot temperature for the to-be-measured hot spot fault, the error influence of the division accuracy of the area information on the temperature is effectively reduced, the accuracy of the temperature diagnosis is improved, and the overall accuracy of the detection result is improved. BRIEF DESCRIPTION OF DRAWINGS
[0033] Figure 1 FIG. 1 is a flowchart of the non-destructive detection method of the hot spot temperature;
[0034] Figure 2 FIG. 2 is a structural schematic diagram of the geometric model of the transformer;
[0035] Figure 3 FIG. 3 is an isotherm distribution diagram under the influence of the hot spot;
[0036] Figure 4 FIG. 4 is a sound pressure distribution diagram of the hot spot fault position I under different heat source powers;
[0037] Figure 5 FIG. 5 is a sound pressure distribution diagram of the hot spot fault position I when the heat source power is 100W;
[0038] Figure 6 FIG. 6 is a sound pressure distribution diagram of the hot spot fault position II when the heat source power is 100W;
[0039] Figure 7 FIG. 7 is a sound pressure distribution diagram of the hot spot fault position III when the heat source power is 100W;
[0040] Figure 8 Figure 9 is a sound pressure distribution graph of a hotspot fault position IV when the power of the heat source is 100W;
[0041] Figure 9 Figure 10 is a graph of an ultrasonic signal received by the ultrasonic sensor E;
[0042] The components represented by the reference numerals in the figures are as follows:
[0043] 1, ultrasonic sensor A; 2, ultrasonic sensor B; 3, ultrasonic sensor C; 4, ultrasonic sensor D; 5, ultrasonic sensor E; 6, ultrasonic sensor F; 7, hotspot fault position I; 8, hotspot fault position II; 9, hotspot fault position III; 10, hotspot fault position IV; 11, transformer winding. DETAILED DESCRIPTION
[0044] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings.
[0045] Embodiments
[0046] The present application provides a winding hotspot temperature nondestructive detection method based on active ultrasonic emission, as shown in Figure 1 The method specifically comprises the following steps:
[0047] S1, establishing an ultrasonic propagation model based on the propagation process of sound waves between different media;
[0048] S2, establishing a geometric model of the transformer, and setting a plurality of ultrasonic sensors on the surface of the box of the geometric model, determining the sound source signal and excitation voltage of the ultrasonic sensors;
[0049] S3, simulating a hotspot fault by setting a point heat source in the winding of the geometric model, simulating the temperature field of the hotspot fault at different positions at different temperatures, and obtaining a temperature field distribution data set;
[0050] S4, based on the internal temperature field distribution of the geometric model of the transformer, using the ultrasonic propagation model to perform active ultrasonic emission, obtaining a sound pressure distribution graph of the ultrasonic wave passing through the temperature field, and obtaining a data set of the sound pressure distribution;
[0051] S5, based on the sound field simulation of the temperature field propagation of the ultrasonic propagation model, using an ultrasonic detection method of one emission and multiple receptions to obtain the sound wave signal, and extracting the effective sound wave signal according to the data set of the sound pressure distribution;
[0052] S6, extracting signal characteristic parameters according to the effective sound wave signal, training a back propagation neural network, establishing an inversion database of hotspot area-hotspot temperature, and detecting and classifying the winding hotspot to be measured through the trained back propagation neural network, to obtain winding hotspot information.
[0053] In step S1, the propagation process of ultrasonic waves between different media inside the transformer is actually a process of forward transmission of waves varying with space and time between fluid and solid, so the establishment of the ultrasonic wave propagation model specifically includes:
[0054] S11, set the wave equation in the fluid, assuming that the transformer oil is a static, uniform and continuous fluid medium, by describing the basic equation of a small acoustic wave of an ideal liquid, the three-dimensional wave equation of acoustic pressure in the fluid is obtained:
[0055]
[0056] p = p(ρ, T)
[0057] Wherein, ρ is the medium density, is a vector differential operator, p is the acoustic pressure change, T is the temperature value, c0 is the static sound speed of the fluid, is the Laplace operator, t is the time.
[0058] S12, set the wave equation in the solid, when the ultrasonic wave propagates in the solid, the description equation is composed of the force balance equation, the geometric deformation equation and the material physical equation, the three-dimensional wave equation of acoustic pressure in the solid medium is obtained:
[0059]
[0060] Wherein, ρ is the medium density, s is the mass point degree, t is the time, λ and μ are the Lami constants of the material.
[0061] S13, set the acoustic-solid coupling boundary condition, according to the propagation process of ultrasonic waves between fluid and solid, based on the direct coupling of structure finite element method and acoustic boundary element method, the boundary condition is set as:
[0062]
[0063] F A = p t n
[0064] Wherein, n is the surface normal direction, p t is the total sound pressure, q d is the acoustic dipole field sound source, u tt is the structure acceleration, F A is the load acting on the structure, ρ is the medium density, c is the sound speed.
[0065] In step S2, the geometric model of the transformer is established, such as Figure 2As shown, the transformer simulation model is simplified and modeled, the transformer oil, winding and box are retained, and it is different from the actual transformer internal winding center placement feature. The internal winding of the geometric model is arranged on one side of the transformer box, the symmetry of the hot spot position and the sensor position is weakened, the difference of the sensor receiving ultrasonic signals under different temperature fields is increased, which is beneficial to the identification of the hot spot position and the hot spot temperature.
[0066] Further, in order to realize the emission and reception of active ultrasonic signals, a plurality of ultrasonic sensors are arranged on the surface of the geometric model box, and the plurality of ultrasonic sensors are used to realize that one ultrasonic sensor can receive the emission signals of other ultrasonic sensors. Through the early sound wave test, it is found that when the sound source signal is a single cycle signal, the sound wave signal received by the sensor is not good, the sound wave amplitude is small and the wave form is complex, and it is difficult to effectively distinguish the sound wave flight time. Therefore, the sound source signal is a continuous wave of 10 cycles, and the amplitude of the excitation voltage is 200V, which can enhance the intensity and recognition degree of the ultrasonic signal.
[0067] In step S3, the hot spot fault is simulated by setting a point heat source in the winding of the geometric model, and the heat source power and heat source position and other parameters are set. The temperature field of the hot spot fault at different positions under different temperatures is simulated, the steady state temperature field of the transformer internal heat balance state is obtained, and the temperature field distribution data set is established.
[0068] By changing the point heat source parameters of the hot spot fault at different positions and different temperatures, the temperature field distribution under different hot spot faults is obtained. Specifically, x hot spot fault positions and y hot spot fault temperatures are set, and x*y types of hot spot fault types are set.
[0069] As a specific embodiment, the method sets 20 types of hot spot fault types, that is, 4 hot spot fault positions and 5 hot spot fault temperatures are set for cross combination. The control variable method is used for comparison and processing, and the distribution characteristics of the temperature field of the transformer when the hot spot fault exists are obtained. The change of the hot spot fault temperature is obtained by changing the heat source power. Taking no hot spot as the reference group (heat source power is 0W), the heat source power setting parameters are 20W, 40W, 60W, 80W and 100W.
[0070] As shown in Figure 3 According to the distribution of the isothermal surface under the influence of the hot spot, when the hot spot exists, the highest temperature is located at the hot spot. With the heat spreading from the hot spot to the surrounding area, the medium temperature in the area near the hot spot increases obviously. With the decrease of the temperature value, the isothermal surface begins to expand from the fault area, and gradually presents the same ring distribution as the winding line cake.
[0071] Furthermore, the resulting temperature field varies depending on the heat source power. Table 1 summarizes the highest winding temperature values within the resulting temperature field based on different hotspot faults.
[0072] Table 1 Summary of the highest temperatures in different hotspot fault temperature fields
[0073]
[0074] As shown in Table 1, the heat source power and the maximum temperature value exhibit a near-linear growth. When the heat source power is the same, the maximum temperature values of the hotspots at locations I, II, and III are equal, while the maximum temperature of the hotspot at location IV is always slightly higher than that at locations I, II, and III, but the difference is not significant.
[0075] In step S4, based on the temperature field distribution inside the geometric model of the transformer, the ultrasonic sensor actively emits ultrasonic signals using the ultrasonic propagation model, and obtains the sound pressure distribution map of the ultrasonic waves passing through the temperature field, thus obtaining a dataset of sound pressure distribution and analyzing the distribution law of sound pressure affected by the temperature field.
[0076] As a specific implementation method, taking the ultrasonic sensor A emitting a sound source signal as an example, the sound pressure distribution map of its sound source signal passing through different hot spot faults is measured and obtained, see [reference]. Figure 4 As shown, when the hotspot fault is at location I, the heat source power is 0W, 20W, 60W, and 100W, and t = t d The sound pressure distribution near the winding is shown in the diagram. When the heat source power is 0W, the sound pressure distribution near the winding exhibits a certain regularity. As the sound wave diffuses away from the sound source, the sound pressure value alternates between positive and negative, and the sound pressure isosurface shows a relatively wide elliptical band distribution. As the local temperature of the winding increases, the band distribution of the sound pressure isosurface becomes curved. When the heat source power increases to 60W, the sound pressure isosurface near the heat source tends to move closer to the heat source. When the heat source power increases to 100W, the temperature field distribution range caused by the heat source increases, and the influence range on the ultrasound further expands. The above indicates that the propagation speed of ultrasound changes under the influence of the heat source temperature distribution.
[0077] See Figures 5-8 As shown, the hotspot faults occur at locations I, II, III, and IV, with a heat source power of 100W and t = t. dThe sound pressure distribution of the sound source signal of the time sensor A near the winding is shown in the figure. It can be seen from the figure that when the hot spot is at position I, the sound pressure isosurface in the four views is obviously curved, and the curved direction is towards the position of the sound source. When the hot spot moves between different positions, it can be found through comparison of the views that the hot spot at the position will affect the sound pressure distribution at the position, but the temperature field of the hot spot at other positions will not cause obvious changes in the sound pressure. Therefore, the distribution rule of the sound pressure is that when the hot spot is far away from the sound source, the influence of the temperature change of the hot spot on the sound pressure distribution will be weakened.
[0078] In step S5, based on the simulation of the sound field propagation in the temperature field according to the ultrasonic wave propagation model, an ultrasonic detection method of one transmission and multiple receptions is used for a plurality of fixed position ultrasonic sensors, so as to obtain all the received sound wave signals, and the effective sound wave signal is extracted according to the data set of the sound pressure distribution;
[0079] In a specific embodiment, taking the ultrasonic sensor A as an example, and using other ultrasonic sensors to receive the ultrasonic wave signals passing through the temperature field, the ultrasonic wave signals are collected and summarized. According to the law of sound wave propagation, with the increase of the propagation time, the received sound wave signals become complex and difficult to distinguish due to repeated reflection and refraction. Therefore, the method intercepts the ultrasonic wave signals within 1.6 ms for waveform analysis.
[0080] As shown in Figure 9 , when the hot spot fault is at position I, the ultrasonic wave signal curve received by the ultrasonic sensor E when the heat source power is 20W, 40W, 60W, 80W and 100W is shown. According to the ultrasonic wave signal collected by the ultrasonic sensor E, the oscillation signal appears within 0.5-0.6ms, and with the increase of the heat source power, the time of the appearance of the sound wave signal increases. This phenomenon also reflects that the increase of the temperature will lead to the decrease of the propagation speed of the sound wave. Through waveform comparison and induction, the ultrasonic wave signal can be divided into three stages: sound wave starting signal, sound wave effective signal and sound wave overlapping signal.
[0081] The sound wave effective signal is after the sound wave starting signal, the starting time and the ending time of the sound wave effective signal have amplitude mutations, the starting time is the time when the rising amplitude of the trough-peak is the largest, the ending time is the time when the falling amplitude of the peak-trough is the largest, and the starting time and the ending time include 10 continuous waves, as shown in the dashed box in Figure 9 . The sound wave starting signal is before the sound wave effective signal, with the increase of the propagation time, the sound wave signal in this stage gradually increases with oscillation, and has high similarity, but the starting amplitude of the sound wave starting signal is small, and the starting time is difficult to determine; the sound wave overlapping signal is after the sound wave effective signal, the sound wave signal in this stage has changed in amplitude compared with the previous stage, and has no obvious regularity.
[0082] In step S6, the signal characteristic parameters are extracted from the acoustic wave effective signal by using a singular value decomposition (SVD) method, the mutation information is extracted through decomposition layers, and the characteristic parameters with high correlation with the hot spot fault are screened out by using a Spearman correlation analysis method; the types of the signal characteristic parameters include peak time, time domain index, and frequency domain index.
[0083] Further, the steps of training the back propagation neural network include:
[0084] The temperature field samples of different hot spot areas are established by cross combination of different heat source powers, and the temperature field samples include signal characteristic parameters of acoustic wave effective signals.
[0085] The area diagnosis model and the temperature diagnosis model of the back propagation neural network are established, and the mapping relationship between the temperature field samples of the hot spot fault and the signal characteristic parameters is fitted and trained.
[0086] The back propagation neural network (BPNN) is composed of an input layer, a hidden layer, and an output layer, and the threshold value and the weight of the neural network are adjusted in a reverse direction according to the error between the actual result and the predicted result by using a gradient descent method until the maximum training number or the output result error meets the accuracy.
[0087] The hot spot fault area is classified by using the area diagnosis model, and the heat source power of a single area is classified by using the temperature diagnosis model corresponding to the single area.
[0088] Further, based on the classification training structure of the back propagation neural network, an inversion database of hot spot area-hot spot temperature is established to realize the inversion of the hot spot fault.
[0089] Further, when an unknown hot spot appears in the internal winding of an actual transformer, the same arrangement strategy of the ultrasonic sensor as the geometric model of the transformer is arranged on the outer wall of the transformer, the ultrasonic signal is emitted to the internal winding of the transformer, the ultrasonic signal is received by other ultrasonic sensors through one-to-many reception, the acoustic wave effective signal is extracted from the received ultrasonic signal, the signal characteristic parameters of the acoustic wave effective signal are extracted, the trained back propagation neural network is used to classify and identify the characteristic parameters, the hot spot fault area is classified according to the area diagnosis model, and the heat source power of a single area is classified by using the temperature diagnosis model, the actual hot spot fault information of the winding to be measured is obtained, and the nondestructive detection of the winding hot spot temperature is realized.
[0090] In addition, the application also provides a winding hot spot temperature nondestructive detection system based on active ultrasonic emission, comprising:
[0091] An ultrasonic wave propagation model establishing module is configured to establish an ultrasonic wave propagation model based on the propagation process of sound waves between different media inside the transformer.
[0092] A transformer model module is configured to establish a geometric model of the transformer, and set a plurality of ultrasonic sensors on the surface of the tank of the geometric model, and determine the sound source signal and excitation voltage of the ultrasonic sensors.
[0093] A temperature field distribution module is configured to simulate hot spot faults by setting point heat sources in the geometric model, simulate the temperature field of the hot spot faults at different temperatures at different positions, and obtain a temperature field distribution dataset.
[0094] An acoustic pressure distribution module is configured to perform active ultrasonic emission using the ultrasonic wave propagation model based on the internal temperature field distribution of the geometric model of the transformer, obtain an acoustic pressure distribution map of the ultrasonic wave passing through the temperature field, and obtain a dataset of the acoustic pressure distribution.
[0095] An acoustic wave signal acquisition module is configured to simulate the acoustic field propagation in the temperature field based on the ultrasonic wave propagation model, acquire acoustic wave signals using an ultrasonic detection method of one emission and multiple receptions, and extract acoustic effective signals according to the dataset of the acoustic pressure distribution.
[0096] An inversion processing module is configured to extract signal characteristic parameters according to the acoustic effective signals, train a back propagation neural network, establish an inversion database of hot spot regions-hot spot temperatures, perform detection and classification on the hot spots to be measured by using the trained back propagation neural network, and obtain winding hot spot information.
[0097] Further, the transformer model module comprises a steel tank, a transformer winding, and transformer oil.
[0098] The hot spot fault positions are uniformly set at the same height position in the upper part of the transformer winding, and are all located outside the transformer winding; the plurality of ultrasonic sensors are all set on the surface of the steel tank outside the transformer, and are located at the same height position as the hot spot faults.
Claims
1. A method for non-destructive detection of hot spot temperature in a winding based on active ultrasonic emission, characterized in that, The method comprises the following steps: S1, establishing an ultrasonic propagation model based on the propagation process of sound waves between different media, specifically including setting up the sound pressure wave equation in fluid and solid and the acoustic-solid coupling boundary condition to establish the ultrasonic propagation model; S2, establishing a geometric model of the transformer, and setting a plurality of ultrasonic sensors on the surface of the geometric model of the tank, determining the sound source signal and excitation voltage of the ultrasonic sensor; S3, simulating hot spot faults by setting point heat sources in the geometric model, simulating the temperature field of hot spot faults at different positions at different temperatures, and obtaining the temperature field distribution data set; S4, based on the internal temperature field distribution of the geometric model of the transformer, using the ultrasonic propagation model to actively emit ultrasonic waves, obtaining the sound pressure distribution diagram of the ultrasonic waves passing through the temperature field, and obtaining the data set of the sound pressure distribution; S5, based on the sound field simulation of the ultrasonic propagation model in the temperature field, using the ultrasonic detection method of one emission and multiple receptions to obtain the sound wave signal, and extracting the sound wave effective signal according to the data set of the sound pressure distribution; S6, extracting signal characteristic parameters according to the sound wave effective signal, training the back propagation neural network, establishing an inversion database of hot spot area-hot spot temperature, and detecting and classifying the hot spot of the winding to be measured through the trained back propagation neural network, to obtain the winding hot spot information. The training of the back propagation neural network comprises: establishing temperature field samples of different hot spot areas and different heat source power cross combinations, the temperature field samples comprising signal characteristic parameters of a plurality of sound wave effective signals; establishing a region diagnosis model and a temperature diagnosis model of the back propagation neural network, and fitting and training the mapping relationship between the temperature field samples and the signal characteristic parameters of the hot spot fault; classifying the hot spot fault area through the region diagnosis model, and classifying the heat source power through the temperature diagnosis model corresponding to a single region.
2. The active ultrasonic emission based non-destructive detection of hot spot temperature in windings method as claimed in claim 1, wherein, The type of the signal characteristic parameters extracted from the sound wave effective signal includes peak time, time domain index and frequency domain index.
3. The active ultrasonic emission based non-destructive detection of hot spot temperature in windings method as claimed in claim 1, wherein, The sound source signal of the ultrasonic sensor is a continuous wave of 10 cycles, and the amplitude of the excitation voltage is 200V.
4. The active ultrasonic emission based non-destructive detection of hot spot temperature in windings method as claimed in claim 1, wherein, The method for extracting the sound wave effective signal from the data set of the sound pressure distribution is that the sound wave effective signal is after the sound wave starting signal, the starting time and the ending time of the sound wave effective signal have amplitude mutation, the starting time is the time when the rising amplitude of the trough-peak is the largest, the ending time is the time when the falling amplitude of the peak-trough is the largest, and the starting time and the ending time include 10 continuous waves.
5. The active ultrasonic emission based non-destructive detection of hot spot temperature in windings method as claimed in claim 1, wherein, The hot spot fault is simulated by setting a point heat source in the geometric model, and the specific method is to set x a hot spot fault position and y a hot spot fault temperature, and a total of x*y hot spot fault types.
6. A system for non-destructive detection of hot spot temperature in a winding based on active ultrasonic emission, characterized in that, comprises: an ultrasonic propagation model establishment module, configured to establish an ultrasonic propagation model based on the propagation process of sound waves between different media in the transformer, specifically including setting up the sound pressure wave equation in fluid and solid and the acoustic-solid coupling boundary condition to establish the ultrasonic propagation model; a transformer model module, configured to establish a geometric model of the transformer, and set a plurality of ultrasonic sensors on the surface of the geometric model of the tank, and determine the sound source signal and excitation voltage of the ultrasonic sensor; The temperature field distribution module is configured to simulate a hot spot fault by setting a point heat source in the geometric model, simulate a temperature field of the hot spot fault at different temperatures at different positions, and obtain a temperature field distribution dataset; The sound pressure distribution module is configured to perform active ultrasonic emission by using an ultrasonic propagation model based on the internal temperature field distribution of the geometric model of the transformer, obtain a sound pressure distribution diagram of the ultrasonic wave passing through the temperature field, and obtain a dataset of the sound pressure distribution; The sound wave signal acquisition module is configured to obtain sound wave signals by using a one-transmit-multiple-receive ultrasonic detection method based on the simulation of the sound field propagating in the temperature field by using the ultrasonic propagation model, and extract sound wave effective signals according to the dataset of the sound pressure distribution; The inversion processing module is configured to extract signal characteristic parameters according to the sound wave effective signals, train a back propagation neural network, establish an inversion database of hot spot regions-hot spot temperatures, detect and classify the hot spot of the to-be-measured winding by using the trained back propagation neural network, and obtain winding hot spot information. The training of the back propagation neural network includes: establishing temperature field samples of different hot spot regions and different heat source power cross combinations, the temperature field samples including signal characteristic parameters of a plurality of sound wave effective signals; establishing a region diagnosis model and a temperature diagnosis model of the back propagation neural network, and fitting and training a mapping relationship between the temperature field samples and the signal characteristic parameters of the hot spot fault; classifying the hot spot fault region by using the region diagnosis model, and classifying the heat source power by using the temperature diagnosis model corresponding to a single region.
7. The active ultrasonic emission based non-destructive probing system for detecting hot spot temperature in windings as claimed in claim 6 wherein, The transformer model module includes a steel tank, a transformer winding, and transformer oil.
8. The active ultrasonic emission based non-destructive probing system for detecting hot spot temperature in windings as claimed in claim 6 wherein, The hot spot fault positions are uniformly set at the same height position in the upper part of the transformer winding and are located on the outer side of the transformer winding.