Load tap changer insulation oil degradation detection method and system based on multi-frequency ultrasound

Abstract

The application belongs to the technical field of ultrasonic detection, and particularly relates to a loaded-tap-changer insulating oil attenuation detection method and system based on multi-frequency ultrasonic waves. The loaded-tap-changer insulating oil attenuation detection method based on multi-frequency ultrasonic waves comprises the following steps: obtaining an initial radius data set of a plurality of bubbles in insulating oil; substituting the initial radius data set into the Minsner equation to obtain a target excitation frequency sequence and a bubble mapping relationship; sweeping frequency to emit an ultrasonic pulse train to the insulating oil according to the target excitation frequency sequence, and receiving a secondary radiation echo signal; transforming the secondary radiation echo signal to a frequency domain to obtain a resonance frequency domain response data; obtaining an abnormal energy loss cumulative value according to the resonance frequency domain response data and the bubble mapping relationship, wherein the abnormal energy loss cumulative value is used to represent a total amount of ultrasonic energy lost due to bubble resonance abnormality; and obtaining a dissolved gas concentration prediction value according to the abnormal energy loss cumulative value and a support vector machine regression model.

Description

Technical Field

[0001] This application belongs to the field of ultrasonic testing technology, specifically relating to a method and system for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound. Background Technology

[0002] On-load tap changers are core components of power transformers for voltage regulation, and the health of their internal insulating oil directly determines the operational safety of the power system. Insulating oil ages and degrades under long-term high voltage and arcing impacts; accurately detecting the degree of degradation of the insulating oil is crucial for preventing sudden equipment failures.

[0003] Traditional detection methods typically rely on extracting oil samples for macroscopic physicochemical analysis. This approach suffers from significant detection lag and is extremely insensitive to microscopic material changes that occur during the early degradation of insulating oil. When insulating oil begins to deteriorate, trace amounts of dissolved gas precipitate and form tiny bubbles of varying sizes. The size distribution and number of these tiny bubbles are important indicators for assessing early oil degradation. However, conventional ultrasonic testing techniques use fixed-frequency or broadband uniform frequency sweep methods. Because the emission frequency cannot precisely match the inherent resonant frequencies of bubbles of different sizes, sound waves passing through the bubbles only cause extremely weak deformation, failing to induce a strong nonlinear resonant response. This results in minimal change in the sound energy received by the testing instrument, making it difficult to determine whether early oil degradation has occurred. Summary of the Invention

[0004] To address the above issues, this application provides a method and system for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound. The method generates a frequency sweep sequence that precisely matches the inherent resonant frequency of each bubble by measuring the bubble size distribution to excite resonance, and quantifies the degree of oil degradation based on the resonant frequency domain response.

[0005] To achieve the above objectives, this application provides a method for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound, comprising: Obtain the initial radius dataset of multiple air bubbles in the insulating oil; Substituting the initial radius dataset into the Minsner equation yields the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation represents the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence represents the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship represents the bubble radius corresponding to each resonant frequency. The ultrasonic pulse train is transmitted to the insulating oil according to the target excitation frequency sequence, and the secondary radiation echo signal is received. The secondary radiation echo signal is transformed to the frequency domain to obtain resonant frequency domain response data, which is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency. Based on the resonant frequency domain response data and the bubble mapping relationship, the cumulative value of abnormal energy loss is obtained, which is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance. Based on the cumulative value of abnormal energy loss and the support vector machine regression model, a predicted value of dissolved gas concentration is obtained, which is used to represent the degree of deterioration of the insulating oil.

[0006] On the one hand, this application provides a multi-frequency ultrasonic-based on-load tap changer insulating oil attenuation detection system, including: The bubble detection module is used to obtain the initial radius dataset of multiple bubbles in the insulating oil. The frequency generation module is used to substitute the initial radius dataset into the Minsner equation to obtain the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation is used to represent the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence is used to represent the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship is used to represent the bubble radius corresponding to each resonant frequency. An ultrasonic transceiver module is used to transmit ultrasonic pulse trains to the insulating oil according to the target excitation frequency sequence and to receive secondary radiation echo signals. The frequency domain transformation module is used to transform the secondary radiation echo signal to the frequency domain to obtain resonant frequency domain response data. The resonant frequency domain response data is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency. An abnormal accumulation module is used to obtain an abnormal energy loss accumulation value based on the resonance frequency domain response data and the bubble mapping relationship. The abnormal energy loss accumulation value is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance. The concentration prediction module is used to obtain a predicted value of dissolved gas concentration based on the cumulative value of abnormal energy loss and the support vector machine regression model. The predicted value of dissolved gas concentration is used to represent the degree of deterioration of the insulating oil.

[0007] The embodiments of this application have the following beneficial effects: In the embodiments of this application, firstly, a quantitative relationship between the bubble radius and the natural resonant frequency is established using the Minsner equation. This allows the embodiments of this application to calculate the natural resonant frequency of each bubble individually based on the measured radius, rather than using a fixed-step frequency sweep method. Simultaneously, the target excitation frequency sequence is arranged in ascending order of radius, with a one-to-one correspondence between the frequency order and the radius order. This ensures that each target excitation frequency corresponds one-to-one with the echo signal during frequency sweep transmission, preventing cross-contamination. In existing technologies, even if ultrasonic echoes are acquired, there is a lack of means to establish a correspondence between frequency and radius. This prevents existing technologies from reverse-engineering the bubble size from the frequency domain response, thus hindering the accurate assessment of the degree of degradation.

[0008] Furthermore, the bubble only resonates strongly under acoustic excitation at a frequency equal to its inherent resonant frequency, generating a secondary radiation echo signal. The resonant frequency domain response data records the amplitude attenuation rate at each target excitation frequency. By analyzing the differences in attenuation rates at different frequencies, frequency channels with abnormally high attenuation can be identified. Since the emitted sound pressure is known and constant at each frequency, the magnitude of the attenuation rate directly reflects the bubble's ability to scatter and absorb the incident sound wave at that frequency. In existing technologies, the emitted energy at each frequency is inconsistent, and the attenuation rate of each frequency channel fluctuates due to differences in emitted energy. The attenuation rates lack comparability, making it impossible to accurately determine whether the abnormal attenuation at a certain frequency is caused by bubble anomalies or transducer frequency response differences.

[0009] Meanwhile, the bubble mapping relationship records the correspondence between each target excitation frequency and the corresponding bubble radius. When a certain frequency is marked as an abnormal frequency, the bubble radius range corresponding to that frequency can be located in reverse through this mapping relationship, completing the accurate inversion from frequency domain response to abnormal bubbles, and then identifying the contribution range of micro water, particle size and acid value corresponding to abnormal bubbles, providing a multi-parameter joint judgment basis for intelligent sensors.

[0010] Furthermore, there is a highly nonlinear mapping relationship between the cumulative value of abnormal energy loss and the concentration of dissolved gas, making it difficult to establish a direct connection through analytical models. In the embodiments of this application, the support vector machine regression model maps the input data to a high-dimensional feature space through the radial basis kernel function, learns and fits the nonlinear mapping relationship in a data-driven manner, maps the acoustic physical quantity of cumulative abnormal energy loss to the predicted value of dissolved gas concentration, and outputs a quantitative assessment result that can directly guide operation and maintenance decisions. Existing technologies rely on empirical thresholds for segmented judgment, which is highly subjective and has poor adaptability, and cannot achieve accurate assessment covering all operating conditions. Moreover, they cannot directly output the dissolved gas concentration index that operation and maintenance personnel care about, resulting in an information gap between the detection results and operation and maintenance decisions. After integrating the above methods, the intelligent sensor can output multi-parameter monitoring data such as micro-water, particle size, acid value, and dissolved gas concentration in situ in real time, realizing a comprehensive assessment and early warning of insulating oil quality. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the first embodiment of this application. Figure 2 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the second embodiment of this application. Figure 3 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the third embodiment of this application. Figure 4 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the fourth embodiment of this application. Figure 5 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the fifth embodiment of this application. Figure 6 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the sixth embodiment of this application. Figure 7 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the seventh embodiment of this application. Figure 8 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the eighth embodiment of this application. Figure 9 This is a flowchart illustrating the on-load tap changer insulating oil attenuation detection method based on multi-frequency ultrasound provided in the ninth embodiment of this application. Figure 10 This is a schematic diagram of the module of the on-load tap changer insulating oil attenuation detection system based on multi-frequency ultrasound provided in the embodiments of this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the technical solution, the present application will be described in detail below with reference to the embodiments. The description in this section is only exemplary and explanatory, and should not be used to limit the scope of protection of the present application in any way.

[0013] In practical applications of insulating oil degradation detection, multi-frequency ultrasonic methods are often used to obtain acoustic information from the oil. Specifically, by emitting ultrasonic waves of multiple frequencies into the oil and receiving the echo signals, acoustic parameters related to the oil's state can be extracted. However, conventional multi-frequency ultrasonic testing typically uses fixed frequency points or arithmetic step sweeps, making it difficult to precisely match the emission frequency with the inherent resonant frequencies of bubbles of different sizes in the oil. This results in the bubbles failing to generate strong nonlinear resonant responses, and the abnormal energy loss information carried in the echo signals is extremely weak, making it difficult to separate effective degradation characteristics from normal physical fluctuations. Furthermore, existing methods lack a complete technical chain for reverse-locating abnormal bubbles from the frequency domain response and transforming microscopic bubble behavior into macroscopic degradation indicators.

[0014] Specifically, the size distribution of bubbles in oil is unknown and random. The radius range of bubbles varies greatly under different deterioration stages and operating conditions, making it impossible to predetermine the frequency sweep range. If a wide-band full-coverage frequency sweep is used, the number of frequency points is enormous, resulting in excessively long detection time and failing to meet real-time monitoring requirements. If a sparse frequency sweep is used, it is easy to miss the resonance frequencies of key bubbles. Therefore, a method is needed that can adaptively generate a frequency sweep sequence based on the actual state of the oil. At the same time, there is a non-linear inverse relationship between the inherent resonance frequency of a bubble and its radius. Small-sized bubbles correspond to high frequencies, and large-sized bubbles correspond to low frequencies. The resonance frequency intervals of bubbles with different radii are uneven. Conventional arithmetic step frequency sweeps may be too dense in the low-frequency range, causing detection redundancy, while in the high-frequency range they may be too sparse, missing resonance points.

[0015] Therefore, how to adaptively generate a frequency sweep sequence that precisely matches the inherent resonant frequency of each bubble based on the actual size distribution of bubbles in the oil to achieve resonant excitation of each bubble, and how to quantify the degree of oil degradation based on the resonant frequency domain response, has become a technical problem that urgently needs to be solved in this field.

[0016] To address the aforementioned technical problems, this application provides a method and system for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound. The method generates a frequency sweep sequence that precisely matches the inherent resonant frequency of each bubble by measuring the bubble size distribution to excite resonance, and quantifies the degree of oil degradation based on the resonant frequency domain response.

[0017] In one embodiment, see Figure 1 This paper provides a method for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound, which includes at least the following steps: Step S100: Obtain the initial radius dataset of multiple air bubbles in the insulating oil.

[0018] Step S120: Substitute the initial radius dataset into the Minsner equation to obtain the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation is used to represent the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence is used to represent the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship is used to represent the bubble radius corresponding to each resonant frequency.

[0019] Step S140: Transmit an ultrasonic pulse train to the insulating oil according to the target excitation frequency sequence and receive the secondary radiation echo signal.

[0020] Step S160: Transform the secondary radiation echo signal to the frequency domain to obtain the resonant frequency domain response data. The resonant frequency domain response data is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency.

[0021] Step S180: Based on the resonance frequency domain response data and the bubble mapping relationship, the cumulative value of abnormal energy loss is obtained. The cumulative value of abnormal energy loss is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance.

[0022] Step S200: Based on the cumulative value of abnormal energy loss and the support vector machine regression model, the predicted value of dissolved gas concentration is obtained. The predicted value of dissolved gas concentration is used to represent the degree of deterioration of the insulating oil.

[0023] It is understandable that the above method can be applied to online monitoring of the insulating oil quality of on-load tap changers. The intelligent sensor can be installed in the insulating oil circulation pipeline or at the drain valve. Specifically, when the intelligent sensor is installed in the insulating oil circulation pipeline, ultrasonic testing is performed as the insulating oil flows through the sensor, and the tested oil continues to flow back into the pipeline without affecting the normal operation of the oil circuit. When the intelligent sensor is installed at the drain valve, it is connected to the drain valve via an adapter, enabling in-situ real-time monitoring without structural modifications to the equipment.

[0024] The intelligent sensor incorporates an ultrasonic transducer. During detection, it emits a series of ultrasonic pulses at a target excitation frequency sequence to the insulating oil flowing through or stationary in the detection area, and receives secondary radiation echo signals. The sensor's signal processing unit transforms the secondary radiation echo signals to the frequency domain, obtaining resonant frequency domain response data. Based on the resonant frequency domain response data and the bubble mapping relationship, it calculates the cumulative abnormal energy loss value. Then, based on the cumulative abnormal energy loss value and a support vector machine regression model, it obtains the predicted dissolved gas concentration value. These data constitute the evaluation indicators for insulating oil quality, guiding operation and maintenance decisions. This intelligent sensor can output monitoring data in situ in real time without offline sampling or laboratory testing, enabling comprehensive evaluation and early warning of insulating oil quality.

[0025] In step S100, the insulating oil is a mineral insulating oil filled inside the on-load tap changer, used to provide electrical insulation and heat dissipation. When the insulating oil ages and degrades under long-term high voltage and arc impact, the hydrocarbon molecular chains in the oil break down, releasing hydrogen and low-molecular-weight hydrocarbon gases. These gases dissolve in the oil to form dissolved gases. Simultaneously, the water content in the oil increases with the moisture absorption of the insulating material, the particulate matter content increases with mechanical wear, and the acid value increases with oxidative deterioration. Water content, particulate matter, and acid value are three core physicochemical parameters for evaluating the quality of insulating oil.

[0026] In the early stages of insulating oil degradation, the size distribution and quantity of gas are important indicators for assessing early oil degradation. The initial radius is the original radius of the bubble when it is not disturbed by an external sound field.

[0027] The initial radius dataset can be obtained by the host computer from the signal acquisition card of the ultrasonic testing equipment. Specifically, the ultrasonic testing equipment emits ultrasonic pulses into the insulating oil through a transducer. After the transducer receives the echo signal, it is converted into a digital signal by the signal acquisition card. The host computer reads the digital signal from the signal acquisition card and extracts the initial radius value of each bubble from the digital signal. The initial radius dataset is composed of multiple initial radius values.

[0028] In step S120, the Minsner equation describes the physical relationship between the natural resonant frequency of a bubble in a fluid medium and the bubble radius, and its expression is:

[0029] in, This is the natural resonant frequency of the bubble. Where is the bubble radius. The adiabatic index of the gas. The hydrostatic pressure of the oil. This represents the density of the oil.

[0030] Each target excitation frequency in the target excitation frequency sequence is equal to the natural resonant frequency of the bubble of the corresponding radius, meaning that the bubble can resonate at that frequency. The bubble mapping relationship records the correspondence between each target excitation frequency and the corresponding bubble radius.

[0031] Specifically, the host computer can read the initial radius value of each bubble from the initial radius dataset obtained in step S100, substitute each initial radius value into the Minsner equation, and calculate the natural resonant frequency of each bubble. Then, the host computer arranges the calculated natural resonant frequencies in ascending order of radius, establishes a mapping relationship between each natural resonant frequency and its corresponding radius value, and sequentially arranges the arranged natural resonant frequencies to generate a target excitation frequency sequence. After calculating the natural resonant frequency of each bubble, the host computer can use the K-means clustering algorithm to cluster the multiple natural resonant frequencies, grouping resonant frequencies with similar values ​​into the same category, and selecting a representative frequency value from each category as the target excitation frequency corresponding to that category. The host computer also obtains the operating frequency range of the sweep frequency acoustic wave generator, discards natural resonant frequencies outside this range, and retains only the frequency values ​​within the operating frequency range to generate the target excitation frequency sequence. The frequency order in the target excitation frequency sequence corresponds one-to-one with the bubble radius order, and is used to sequentially output each target excitation frequency during subsequent sweep frequency transmission.

[0032] In step S140, frequency sweep transmission refers to controlling the frequency sweep acoustic generator to output ultrasonic pulse trains at each target excitation frequency sequentially, starting from the first target excitation frequency, according to the frequency order in the target excitation frequency sequence. The electrical signal output by the frequency sweep acoustic generator is amplified by a power amplifier and drives the ultrasonic transducer. The ultrasonic transducer converts the electrical signal into an acoustic signal and transmits it to the insulating oil. After the ultrasonic pulse train at each target excitation frequency is transmitted, the ultrasonic transducer switches to receiving mode, waits to receive the echo signal at that frequency, and then switches to the next frequency to continue transmission. After the ultrasonic pulse train is transmitted to the insulating oil, if a certain target excitation frequency is exactly equal to the natural resonant frequency of a certain bubble, the bubble resonates at that frequency. The resonance causes the bubble to produce volume pulsation, and the pulsating bubble acts as a secondary sound source, radiating sound waves outward. This sound wave signal is the secondary radiation echo signal. The ultrasonic transducer receives this secondary radiation echo signal, and after the signal acquisition card converts the analog signal into a digital signal, the host computer reads and stores the digital signal.

[0033] Specifically, when calculating the frequency step interval, the difference between two adjacent target excitation frequencies in the target excitation frequency sequence can be calculated:

[0034] in, and These are the nth values ​​in the target excitation frequency sequence. The and the first The frequency of each target stimulus This represents the frequency step interval. The duration of the ultrasonic pulse train at each target excitation frequency is three times the duration of a single cycle at that frequency.

[0035] This ensures that each bubble is excited by the same number of acoustic cycles. After the ultrasonic pulse train is driven by the power amplifier to be emitted into the insulating oil, the bubble resonates at its natural resonant frequency. The resonant bubble then emits a secondary acoustic signal, which is the secondary radiation echo signal. The transducer receives this secondary radiation echo signal, which is converted into a digital signal by the signal acquisition card and then read and stored by the host computer.

[0036] In step S160, the host computer performs time-frequency transformation processing on the digital signal of the secondary radiation echo signal stored in step S140, converting the time-domain signal into a frequency-domain signal to obtain the frequency-domain signal.

[0037] The host computer extracts the echo energy value corresponding to each target excitation frequency from the frequency domain signal, and determines the amplitude attenuation rate at that frequency based on the echo energy value. The host computer associates and stores each target excitation frequency with its corresponding amplitude attenuation rate, forming resonant frequency domain response data. The resonant frequency domain response data records the amplitude attenuation rate at each target excitation frequency.

[0038] In step S180, the host computer acquires the resonant frequency domain response data obtained in step S160. The resonant frequency domain response data records the amplitude attenuation rate at each target excitation frequency. The host computer also acquires the bubble mapping relationship obtained in step S120. The bubble mapping relationship records the correspondence between each target excitation frequency and the corresponding bubble radius.

[0039] The host computer identifies anomalous frequencies with abnormal amplitude attenuation rates based on the amplitude attenuation rate at each target excitation frequency in the resonant frequency domain response data. Then, according to the bubble mapping relationship, it locates the anomalous frequencies to the corresponding bubble radius range and calculates the sum of the acoustic scattering cross-sectional areas of all bubbles within that radius range to obtain the cumulative value of abnormal energy loss. This cumulative value of abnormal energy loss represents the total ultrasonic energy lost due to abnormal bubble resonance. For example, the cumulative value of abnormal energy loss obtained after accumulation is 42,000 microjoules.

[0040] In step S200, the support vector machine regression model is a machine learning regression algorithm based on the principle of minimizing structural risk, used to establish the mapping relationship between the cumulative value of abnormal energy loss and the concentration of dissolved gas.

[0041] During the pre-training of the support vector machine regression model, the host computer extracts multiple sets of cumulative abnormal energy loss values ​​and corresponding measured dissolved gas concentration values ​​from the historical monitoring database. Each set of cumulative abnormal energy loss values ​​and its corresponding measured dissolved gas concentration value is treated as a sample data set, resulting in an initial sample set. The host computer performs feature normalization processing on the initial sample set, mapping the cumulative abnormal energy loss values ​​and measured dissolved gas concentration values ​​to a uniform range of 0 to 1, obtaining a standard sample set. The host computer then randomly splits the standard sample set into a training set and a test set at a ratio of 7:3.

[0042] The host computer inputs the training set into the support vector machine regression model and uses a grid search method to traverse the preset parameter space. The preset parameter space includes a penalty coefficient ranging from 0.1 to 100 and a kernel parameter ranging from 0.01 to 10. The mean squared error (MSE) of each parameter combination is calculated using five-fold cross-validation. The parameter combination with the smallest MSE is selected as the optimal parameter combination, determining the penalty coefficient and kernel parameter of the radial basis function. For example, 2000 sets of sample data are extracted from the historical monitoring database. After grid search and five-fold cross-validation, the penalty coefficient C is determined to be 10 and the kernel parameter γ to be 0.5.

[0043] The host computer trains a support vector machine regression model using the training set and the determined optimal parameter combination. It then validates the trained model using a test set to evaluate its prediction accuracy. The host computer deploys the trained and validated support vector machine regression model locally or in the cloud.

[0044] The host computer obtains the cumulative value of abnormal energy loss obtained in step S180, inputs the cumulative value of abnormal energy loss into the pre-trained support vector machine regression model, and outputs the predicted value of dissolved gas concentration after calculation by the support vector machine regression model.

[0045] Furthermore, the host computer can also statistically analyze the acoustic scattering cross-sectional area contribution of bubbles of different sizes within the bubble radius range corresponding to the abnormal frequency. This value is then mapped to at least one of the following through a pre-established mapping relationship: micro-moisture prediction, particle size prediction, and acid value prediction. The micro-moisture prediction indicates the micro-moisture content in the insulating oil, the particle size prediction indicates the particulate matter concentration, and the acid value prediction indicates the acid value level. Micro-moisture, particle size, and acid value are three core physicochemical parameters for evaluating the quality of insulating oil, and together with the dissolved gas concentration prediction, they constitute a comprehensive evaluation index set for insulating oil quality.

[0046] The predicted dissolved gas concentration is used to represent the current degree of degradation of the insulating oil. Specifically, when the insulating oil ages and degrades under long-term high voltage and arc impact, the hydrocarbon molecular chains in the oil break, releasing hydrogen and low-molecular-weight hydrocarbon gases. These gases dissolve in the oil to form dissolved gases, and the concentration of dissolved gases increases with the degree of degradation. When the dissolved gases reach saturation, the supersaturated gases precipitate from the oil to form microbubbles. Therefore, the size distribution and number of bubbles directly reflect the dissolved gas content level in the oil, and the level of dissolved gas concentration directly reflects the severity of oil degradation. The cumulative abnormal energy loss value is the total ultrasonic energy lost due to abnormal bubble resonance; that is, this value indirectly reflects the degree of bubble abnormality in the oil, which is determined by the dissolved gas concentration. Therefore, the predicted dissolved gas concentration output by the support vector machine regression model can be used to represent the current degree of degradation of the insulating oil.

[0047] In this embodiment, firstly, a quantitative relationship between the bubble radius and its natural resonant frequency is established using the Minsner equation. This allows the embodiment to calculate the natural resonant frequency of each bubble individually based on the measured radius, rather than using a fixed-step frequency sweep method. Simultaneously, the target excitation frequency sequence is arranged in ascending order of radius, with a one-to-one correspondence between the frequency order and the radius order. This ensures that each target excitation frequency corresponds one-to-one with the echo signal during frequency sweep transmission, preventing cross-contamination. Existing technologies, even when acquiring ultrasonic echoes, lack a means to establish a mapping between frequency and radius. This prevents existing technologies from reverse-engineering the bubble size from the frequency domain response, thus hindering accurate assessment of the degree of degradation.

[0048] Furthermore, the bubble only resonates strongly under acoustic excitation at a frequency equal to its inherent resonant frequency, generating a secondary radiation echo signal. The resonant frequency domain response data records the amplitude attenuation rate at each target excitation frequency. By analyzing the differences in attenuation rates at different frequencies, frequency channels with abnormally high attenuation can be identified. Since the emitted sound pressure is known and constant at each frequency, the magnitude of the attenuation rate directly reflects the bubble's ability to scatter and absorb the incident sound wave at that frequency. In existing technologies, the emitted energy at each frequency is inconsistent, and the attenuation rate of each frequency channel fluctuates due to differences in emitted energy. The attenuation rates lack comparability, making it impossible to accurately determine whether the abnormal attenuation at a certain frequency is caused by bubble anomalies or transducer frequency response differences.

[0049] Meanwhile, the bubble mapping relationship records the correspondence between each target excitation frequency and the corresponding bubble radius. When a certain frequency is marked as an abnormal frequency, the bubble radius range corresponding to that frequency can be located in reverse through this mapping relationship, completing the accurate inversion from frequency domain response to abnormal bubbles, and then identifying the contribution range of micro water, particle size and acid value corresponding to abnormal bubbles, providing a multi-parameter joint judgment basis for intelligent sensors.

[0050] Furthermore, there is a highly nonlinear mapping relationship between the cumulative value of abnormal energy loss and the concentration of dissolved gas, making it difficult to establish a direct connection through analytical models. In this embodiment, the support vector machine regression model maps the input data to a high-dimensional feature space through the radial basis function kernel function, learns and fits this nonlinear mapping relationship in a data-driven manner, maps the acoustic physical quantity of cumulative abnormal energy loss to the predicted value of dissolved gas concentration, and outputs a quantitative assessment result that can directly guide operation and maintenance decisions. Existing technologies rely on empirical thresholds for segmented judgment, which is highly subjective and has poor adaptability, failing to achieve accurate assessment covering all operating conditions, and cannot directly output the dissolved gas concentration index that operation and maintenance personnel care about, resulting in an information gap between the detection results and operation and maintenance decisions. After integrating the above methods, the intelligent sensor can output multi-parameter monitoring data such as micro-water, particle size, acid value, and dissolved gas concentration in situ in real time, realizing a comprehensive assessment and early warning of insulating oil quality.

[0051] In one embodiment, see Figure 2 Step S100 includes: Step S101: Collect the ultrasonic echo signal of the insulating oil in a resting state.

[0052] Step S102: Perform Hilbert transform on the echo signal to obtain the envelope curve.

[0053] Step S103: Determine the backscattering sound pressure of each bubble based on the peak position of the envelope curve.

[0054] Step S104: Calculate the acoustic scattering cross-sectional area of ​​each bubble based on the backscattering sound pressure of each bubble.

[0055] Step S105: Perform regression processing on the acoustic scattering cross-sectional area of ​​each bubble to obtain the initial radius value of each bubble.

[0056] Step S106: Generate an initial radius dataset based on multiple initial radius values.

[0057] In step S101, the ultrasonic echo signal of the insulating oil in a resting state is acquired. The resting state refers to the state where the insulating oil has been sufficiently settled after the equipment has stopped operating, and the air bubbles are in a relatively stable distribution. The ultrasonic echo signal is the sound wave signal that returns to the transducer after the ultrasonic transducer emits an ultrasonic pulse into the insulating oil, propagates through the oil medium, is scattered by air bubbles, and then receives the signal. This signal is received by the ultrasonic transducer, converted into an electrical signal, and then converted into a digital signal by the signal acquisition card, which is then read and stored by the host computer. The ultrasonic transducer is a device used to convert electrical signals into sound waves and emit them into the insulating oil, while simultaneously receiving echo signals and converting the sound waves back into electrical signals. The signal acquisition card is an analog-to-digital converter used to convert analog signals into digital signals; its sampling rate determines the time resolution of the digital signal. The host computer is a computing device used to execute signal processing and control algorithms; specifically, it can be an industrial control computer, a computer, or an embedded processor. In this embodiment, the host computer reads the digital signal from the signal acquisition card and then executes all subsequent signal processing steps.

[0058] In step S102, the host computer performs a Hilbert transform on the echo signal to obtain the envelope curve. Let the original echo signal be , and the host computer performs a Hilbert transform on it to obtain:

[0059] Superior mechanism structural analysis signal:

[0060] Where j is the imaginary unit. The host computer extracts the magnitude of the analytical signal to obtain the envelope curve:

[0061] The envelope curve is the contour line of the amplitude of the echo signal changing over time, reflecting the attenuation trend of sound wave energy on the time axis. Each peak corresponds to an echo event of a bubble.

[0062] In step S103, the host computer determines the backscattered sound pressure of each bubble based on the peak position of the envelope curve. The host computer performs peak detection on the envelope curve, identifying local maxima as peaks. The host computer reads the peak amplitude of each peak and converts it into a sound pressure value, i.e., the backscattered sound pressure of the bubble, according to the pre-calibrated curve of the ultrasonic transducer. Backscattered sound pressure is the sound pressure value generated by the bubble backscattering incident sound waves.

[0063] The host computer also determines the spatial position of each bubble based on the time difference between two adjacent wave peaks. Let the time coordinates of two adjacent wave peaks be... and The time difference is Multiplied by the speed of sound Divide by 2 to get the first... The distance between each bubble and the transducer:

[0064] in For the first The distance between each bubble and the transducer For the speed of sound, and These are the time coordinates of two adjacent wave peaks. For example, when the time difference between two adjacent wave peaks... When the time interval is 5 microseconds and the speed of sound in the insulating oil is c = 1400 m / s, the distance between the bubble and the transducer can be calculated using the formula. The distance is 3 millimeters. After traversing all the peaks, the distance information of multiple bubbles constitutes a three-dimensional spatial distribution coordinate matrix of each bubble in the oil.

[0065] The distance information of multiple bubbles constitutes the spatial distribution of each bubble in the oil. This spatial distribution information corresponds one-to-one with the backscattered sound pressure, and is used in the subsequent step S104 to associate the sound pressure data of the bubble when calculating the acoustic scattering cross-section of each bubble.

[0066] In step S104, the host computer calculates the acoustic scattering cross-sectional area of ​​each bubble based on its backscattered sound pressure and spatial position. The host computer obtains the backscattered sound pressure of each bubble determined in step S103 and the distance between each bubble and the transducer. Sound waves experience propagation attenuation in oil, resulting in a difference between the received backscattered sound pressure and the actual scattered sound pressure, which is related to the sound wave propagation distance. The host computer compensates for the propagation attenuation of the backscattered sound pressure based on the distance between each bubble and the transducer, obtaining the compensated backscattered sound pressure. The propagation attenuation compensation is calculated by multiplying the backscattered sound pressure by a distance square factor, i.e.:

[0067] in, The backscattered sound pressure after compensation. The backscattered sound pressure determined in step S103, For the first The distance between each bubble and the transducer. The host computer obtains the incident sound pressure value, which is the initial sound pressure value when the ultrasonic transducer emits an ultrasonic pulse. The host computer substitutes the compensated backscattered sound pressure of each bubble and the incident sound pressure into the formula to calculate the acoustic scattering cross-section:

[0068] in, For acoustic scattering cross-section, The backscattered sound pressure after compensation. The incident sound pressure is denoted as . The acoustic scattering cross-section is a physical quantity characterizing the ability of a bubble to scatter incident sound waves. Through this propagation attenuation compensation, the influence of different sound wave propagation distances on the backscattered sound pressure of each bubble is eliminated, allowing the calculated acoustic scattering cross-section to accurately reflect the scattering characteristics of each bubble.

[0069] As another calculation method, the host computer can also substitute the backscattered sound pressure into the spherical wave scattering integral equation to calculate the acoustic scattering cross-section. The spherical wave scattering integral equation is as follows:

[0070] in, The scattering amplitude, The scattering angle is denoted by . This equation is applicable to the calculation of the scattering cross-section when the sound field scattered by the bubble satisfies the far-field approximation condition of the spherical wave.

[0071] In step S105, the host computer performs regression processing on the acoustic scattering cross-section of each bubble to obtain the initial radius value of each bubble. The host computer obtains the acoustic scattering cross-section of each bubble calculated in step S104 and performs regression processing on the acoustic scattering cross-section using a random forest algorithm. The physical model used for regression processing is the Rayleigh scattering model. The Rayleigh scattering model describes the relationship between the acoustic scattering cross-section and the bubble radius, and its expression is:

[0072] in, For acoustic scattering cross-section, Where is the bubble radius. The resonant frequency of the bubble. The frequency of the incident sound wave, This represents the damping coefficient. The host computer substitutes the acoustic scattering cross-section of each bubble into the Rayleigh scattering model, and uses the random forest algorithm to adjust the bubble radius in the model. Regression fitting is performed. The random forest algorithm constructs multiple decision trees, combines the fitting results of each decision tree, and outputs the initial radius value of each bubble. The initial radius is the original radius value of the bubble when it is not disturbed by the external sound field.

[0073] As another calculation method, the host computer can also substitute the backscattered sound pressure into the spherical wave scattering integral equation to calculate the acoustic scattering cross-section. The spherical wave scattering integral equation is as follows:

[0074] in, The scattering amplitude, The scattering angle is denoted by . This equation is applicable to the calculation of the scattering cross-section when the sound field scattered by the bubble satisfies the far-field approximation condition of the spherical wave.

[0075] In step S106, the host computer generates an initial radius dataset based on multiple initial radius values. The host computer collects the initial radius values ​​of all bubbles obtained in step S105 and combines these initial radius values ​​into a single dataset, which is the initial radius dataset. This initial radius dataset records the initial radius values ​​of all detected bubbles in the insulating oil and is used in subsequent step S120.

[0076] In this embodiment, firstly, the envelope curve is extracted from the ultrasonic echo signal using Hilbert transform, so that each peak corresponds to an echo event of a bubble, achieving separation of the echo signals of each bubble. Secondly, this embodiment determines the backscattered sound pressure of each bubble based on the peak position of the envelope curve, and calculates the acoustic scattering cross-section based on the ratio of the backscattered sound pressure to the incident sound pressure, eliminating the influence of emitted energy fluctuations on scattering measurements. Finally, this embodiment uses the Rayleigh scattering model and random forest algorithm to regress the acoustic scattering cross-section to the initial bubble radius value, generating an initial radius dataset from multiple initial radius values.

[0077] Existing technologies typically perform amplitude analysis or thresholding directly on the raw echo signal, failing to separate the independent scattering contribution of each bubble from the aliased signal. This results in insufficient accuracy in obtaining the bubble size distribution and neglects the influence of sound wave propagation distance on backscattered sound pressure. This embodiment ensures the accuracy of the initial radius dataset through the aforementioned methods, providing a reliable data foundation for the subsequent generation of the target excitation frequency sequence, thereby ensuring that each bubble can be accurately excited by its inherent resonant frequency during frequency sweeping transmission.

[0078] In one embodiment, see Figure 3 Step S160 includes: Step S161: Apply the Hanning window function to the secondary radiation echo signal to obtain the windowed signal.

[0079] Step S162: Perform a fast Fourier transform on the windowed signal to obtain the frequency domain amplitude spectrum.

[0080] Step S163: Extract the peak amplitude at each target excitation frequency from the frequency domain amplitude spectrum.

[0081] Step S164: Determine the amplitude attenuation rate at each target excitation frequency based on each peak amplitude.

[0082] Step S165: The resonant frequency domain response data is composed of the excitation frequency of each target and the corresponding amplitude attenuation rate.

[0083] In step S161, the host computer applies a Hanning window function to the secondary radiation echo signal to obtain a windowed signal. The secondary radiation echo signal is a finite-length discrete digital signal with truncation abrupt changes at both ends; direct frequency domain transformation will introduce spectral leakage. The Hanning window function is a window function used to suppress spectral leakage, and its expression is:

[0084] in, For the Hanning window function in the th The value of each sampling point, For the length of the window, The sampling point number and The host computer multiplies each sampling point of the secondary radiation echo signal with the value of the corresponding point of the Hanning window function to obtain the windowed signal. For example, the length of the secondary radiation echo signal is 1024 sampling points, the window length N of the Hanning window function is 1024, and the value of n ranges from 0 to 1023.

[0085] In step S162, the host computer performs a Fast Fourier Transform on the windowed signal to obtain the frequency domain amplitude spectrum. Let the windowed signal be... Its length is Then the discrete Fourier transform formula is:

[0086] in, For the first The frequency domain complex number results at each frequency point For windowing signals in the first The value of each sampling point, For signal length, For frequency point number, The imaginary unit is used. The host computer calculates the modulus value at each frequency point. The frequency domain amplitude spectrum is obtained.

[0087] In step S163, the host computer acquires each target excitation frequency value in the target excitation frequency sequence, locates the corresponding frequency point in the frequency domain amplitude spectrum, reads the amplitude value at that frequency point, and obtains the peak amplitude at the target excitation frequency. .

[0088] In step S164, the host computer obtains the initial emitted sound pressure corresponding to the target excitation frequency recorded in step S140. Substituting the peak amplitude and initial emitted sound pressure into the formula, the amplitude attenuation rate is calculated:

[0089] in, Target excitation frequency Amplitude attenuation rate at that point The initial emitted sound pressure, Target excitation frequency The peak amplitude at that frequency. The amplitude attenuation rate is a parameter that reflects the degree of energy attenuation when ultrasound propagates in an oil medium. The larger the value, the more ultrasound energy is absorbed by the oil at that frequency.

[0090] In step S165, the host computer sets the excitation frequency of each target. Its corresponding amplitude attenuation rate The data is stored in a correlation to establish a correspondence between frequency and attenuation rate; this correspondence constitutes the resonant frequency domain response data. The resonant frequency domain response data records the energy attenuation of the ultrasonic wave after passing through the insulating oil at each target excitation frequency, which is used to identify abnormal frequencies in subsequent step S180.

[0091] In this embodiment, a Hanning window function is first applied to the secondary radiation echo signal to eliminate abrupt cutoffs at both ends of the signal, suppress spectral leakage, and make the frequency domain analysis results more accurate. Then, a Fast Fourier Transform (FFT) is used to convert the time-domain echo signal to the frequency domain, obtaining the frequency domain amplitude spectrum, which clearly presents the energy distribution of each frequency component. Next, the peak amplitude at each target excitation frequency is extracted from the frequency domain amplitude spectrum, and the difference between the peak amplitude and the initial emitted sound pressure is calculated to obtain the amplitude attenuation rate at each target excitation frequency. The resonant frequency domain response data is composed of each target excitation frequency and its corresponding amplitude attenuation rate. Existing technologies typically perform amplitude analysis directly on the echo signal in the time domain, failing to distinguish the attenuation differences of different frequency components, and do not perform window function processing on the signal, resulting in insufficient accuracy in frequency domain analysis due to spectral leakage. This embodiment achieves accurate conversion from the time-domain echo signal to frequency domain attenuation characteristics through the above methods, providing a reliable data foundation for subsequent identification of abnormal frequencies.

[0092] In one embodiment, see Figure 4 Step S180 includes: Step S181: Calculate the ratio of amplitude attenuation rate to initial emitted sound pressure at each target excitation frequency.

[0093] Step S182: Mark the target excitation frequency whose ratio exceeds the preset attenuation ratio threshold as an abnormal frequency.

[0094] Step S183: Locate the bubble radius range corresponding to the abnormal frequency through the bubble mapping relationship.

[0095] Step S184: Accumulate the acoustic scattering cross-sectional area of ​​all bubbles within the bubble radius range corresponding to the abnormal frequency to obtain the cumulative value of abnormal energy loss.

[0096] In step S181, the host computer acquires the resonant frequency domain response data obtained in step S165, which records the amplitude attenuation rate corresponding to each target excitation frequency. The host computer also acquires the initial transmitted sound pressure level corresponding to each target excitation frequency recorded in step S140. The host computer divides the amplitude attenuation rate at each target excitation frequency by its corresponding initial transmitted sound pressure level to obtain the normalized attenuation ratio at that target excitation frequency.

[0097] in, Target excitation frequency The normalized decay ratio under the following conditions Target excitation frequency The amplitude attenuation rate below, Target excitation frequency The initial emitted sound pressure level is set below the threshold. This ratio eliminates the influence of differences in emitted sound pressure level on the absolute value of the attenuation rate, making the attenuation levels of different frequency channels comparable.

[0098] In step S182, the preset attenuation ratio threshold is a critical ratio set statistically after ultrasonic testing of multiple normal oil samples. The tuning method is as follows: multiple known normal insulating oil samples are collected, and testing is performed according to steps S100 to S160. The normalized attenuation ratio of each normal sample at each target excitation frequency is calculated. The upper quartile of the ratio distribution of all normal samples at each frequency channel is multiplied by 1.5 as the preset attenuation ratio threshold for that frequency channel. The host computer calculates the normalized attenuation ratio at each target excitation frequency. The preset attenuation ratio threshold corresponding to this frequency channel If a comparison is made, If the target excitation frequency is abnormal, then the excitation frequency is marked as an abnormal frequency. For example, the preset attenuation ratio threshold is three per thousand.

[0099] In step S183, the host computer obtains the bubble mapping relationship obtained in step S120. The bubble mapping relationship records the correspondence between each target excitation frequency and the corresponding bubble radius. This correspondence is a lookup table, where each row records a target excitation frequency value and its corresponding bubble radius value. The host computer uses the target excitation frequency marked as an abnormal frequency as the lookup key to perform a matching query in the bubble mapping relationship table, finds the bubble radius value corresponding to that frequency, and uses this radius value as the bubble radius range corresponding to the abnormal frequency.

[0100] In step S184, the host computer obtains the acoustic scattering cross-sectional area of ​​each bubble calculated in step S104. and their corresponding bubble radii All bubbles whose radii fall within the range corresponding to the anomalous frequency are selected, and the acoustic scattering cross-sections of these bubbles are summed:

[0101] in, This represents the cumulative value of abnormal energy loss. The range of bubble radii corresponding to the abnormal frequency. radius The acoustic scattering cross-sectional area of ​​the bubble. This accumulated value is the cumulative value of abnormal energy loss, used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance.

[0102] In this embodiment, the influence of differences in emitted sound pressure on the attenuation rate is eliminated by calculating the normalized ratio of the amplitude attenuation rate to the initial emitted sound pressure, making the attenuation levels of different frequency channels comparable. Then, the normalized ratio is compared with a preset attenuation ratio threshold. Target excitation frequencies with ratios exceeding the threshold are marked as abnormal frequencies, achieving automatic identification of abnormal frequencies. The preset attenuation ratio threshold is set after statistical analysis of normal oil samples, ensuring statistical basis for the threshold. Next, a matching query is performed using the abnormal frequency as the lookup key through bubble mapping relationships, reversing the abnormal information in the frequency domain to the bubble radius domain, obtaining the bubble radius interval corresponding to the abnormal frequency. Finally, the acoustic scattering cross-sectional area of ​​all bubbles within this radius interval is accumulated, aggregating the microscopic bubble scattering behavior into a macroscopic energy index. Existing technologies lack reverse localization methods from frequency domain response to bubble radius intervals, leaving detection results at a qualitative level. This embodiment achieves accurate inversion from frequency domain anomalies to bubble radius intervals through the above methods, quantifying abnormal energy loss into a numerically assessable index.

[0103] In one embodiment, see Figure 5 Step S140 includes: Step S141: Set the difference between adjacent frequencies in the target excitation frequency sequence as the sweep frequency step interval.

[0104] Step S142: According to the frequency sweep step interval, output the ultrasonic pulse train of each target excitation frequency to the insulating oil in sequence.

[0105] In step S141, the host computer acquires the target excitation frequency sequence generated in step S120. This sequence contains multiple target excitation frequency values ​​arranged in ascending order. The host computer calculates the difference between two adjacent target excitation frequencies in the sequence to obtain the frequency step interval.

[0106] in, For frequency step interval, and These are the nth values ​​in the target excitation frequency sequence. The and the first There are several target excitation frequencies. The step interval is not a fixed value; it changes according to the actual difference between adjacent frequencies in the target excitation frequency sequence, ensuring that each target excitation frequency is sequentially swept and emitted without missing any resonant frequency point of the bubble.

[0107] When the target excitation frequency sequence is an arithmetic sequence, that is... The host computer directly uses this fixed value as the frequency sweep step interval. For example, when the target excitation frequency sequence is an arithmetic sequence from 20kHz to 50kHz, the difference between adjacent frequencies is 500Hz, then the frequency sweep step interval is 500Hz.

[0108] In step S142, the host computer controls the sweep frequency acoustic wave generator to start from the first frequency value of the target excitation frequency sequence and increase the output frequency according to the sweep frequency step interval calculated in step S141. Each time a frequency value is output, the generator pauses for one pulse train duration, during which an ultrasonic pulse train of that frequency is emitted into the insulating oil. After the emission is completed, the generator switches to the next frequency value until all target excitation frequencies in the sequence have been emitted.

[0109] In this embodiment, the difference between adjacent frequencies in the target excitation frequency sequence is first set as the frequency sweep step interval. This step interval changes with the actual difference between adjacent frequencies in the target excitation frequency sequence to ensure that each target excitation frequency is swept and emitted in sequence, without missing any resonant frequency point of the bubble.

[0110] Then, ultrasonic pulse trains at each target excitation frequency are sequentially output to the insulating oil according to the frequency sweep step interval. The duration of each ultrasonic pulse train at each target excitation frequency is three times the duration of a single cycle of the corresponding frequency. This duration ensures that each bubble is excited by the same number of acoustic waves, making the echo signals of each frequency channel comparable. It can be understood that bubbles resonate under acoustic excitation at their inherent resonant frequency, and the establishment of resonance requires a certain number of excitation cycles. If the number of excitation cycles is too small, the bubble has not fully established resonance, and the intensity of the secondary radiation echo signal is insufficient, making effective detection impossible. If the number of excitation cycles is too large, continuous energy injection may cause the bubble to rupture or produce nonlinear effects, interfering with the accuracy of the detection results. Three times the duration of a single cycle is an experimentally verified preferred value. For example, when the target excitation frequency is 50kHz, its single cycle duration is 20μs, and three times that is 60μs. This duration is sufficient for the bubble to establish stable resonance while avoiding nonlinear interference caused by excessive energy injection. The duration setting corresponds one-to-one with the frequency, meaning that the duration of the pulse train at each frequency varies with that frequency, ensuring that each bubble is excited by the same number of sound waves at its own inherent frequency, thus making the echo signals of each frequency channel comparable.

[0111] Existing technologies employ fixed-step frequency sweeping, which, due to its fixed and typically large step size, easily skips the resonant frequencies of some bubbles. Furthermore, the identical transmission duration at each frequency leads to inconsistent excitation cycle numbers across different frequencies, resulting in a lack of comparability in the echo signals. This embodiment achieves precise frequency sweeping of the inherent resonant frequencies of each bubble through the aforementioned methods, providing a prerequisite for accurate extraction of the attenuation rate of each frequency channel and anomaly identification.

[0112] In one embodiment, see Figure 6 Step S140 includes: Step S143: Obtain the frequency response compensation coefficient at each target excitation frequency.

[0113] Step S144: Generate the sound pressure amplitude modulation curve based on the frequency response compensation coefficient.

[0114] Step S145: Control the sound pressure amplitude of the ultrasonic pulse train at each target excitation frequency according to the sound pressure amplitude modulation curve so that the sound pressure amplitude at each target excitation frequency is constant.

[0115] In step S143, the swept-frequency acoustic wave generator is a signal source device that outputs an electrical signal, and the ultrasonic transducer is a conversion device that converts the electrical signal into an acoustic signal and emits it into the insulating oil. The electrical signal output by the swept-frequency acoustic wave generator is amplified by a power amplifier and drives the ultrasonic transducer, which then converts the electrical signal back into an acoustic signal for output. The electroacoustic conversion efficiency of the ultrasonic transducer varies at different frequencies; that is, the sound pressure amplitude produced by applying the same electrical signal amplitude differs at different frequencies. The host computer stores the joint frequency response compensation coefficients of the swept-frequency acoustic wave generator and the ultrasonic transducer at each frequency. These joint frequency response compensation coefficients are obtained through factory calibration or pre-testing and are used to compensate for the frequency response differences throughout the entire transmission link from the electrical signal input to the acoustic wave output. Based on the current target excitation frequency, the host computer reads the compensation value corresponding to that frequency from the stored joint frequency response compensation coefficients and uses it as the frequency response compensation coefficient for that target excitation frequency.

[0116] In step S144, the host computer obtains the frequency response compensation coefficient corresponding to each target excitation frequency, and plots each target excitation frequency and its corresponding frequency response compensation coefficient as a continuous curve in the frequency-amplitude coordinate system. This curve is the sound pressure amplitude modulation curve. The sound pressure amplitude modulation curve reflects the amount of modulation required to make the output sound pressure at each frequency consistent.

[0117] In step S145, when the host computer outputs the ultrasonic pulse train for each target excitation frequency, it modulates the amplitude of the input electrical signal accordingly based on the compensation value corresponding to that frequency on the sound pressure amplitude modulation curve. For frequencies with weak transducer response, the amplitude of the input electrical signal is increased for compensation; for frequencies with strong transducer response, the amplitude of the input electrical signal is decreased for suppression, ensuring that the actual output sound pressure amplitude at each frequency remains consistent. A constant sound pressure amplitude ensures that the attenuation rate of each frequency channel is determined solely by the degree of absorption of the sound waves by the oil, eliminating the influence of differences in emitted sound pressure on the attenuation rate and making the attenuation rates of each frequency channel comparable.

[0118] In one embodiment, see Figure 7 After step S200, the following steps are also included: Step S220: Compare the predicted value of dissolved gas concentration with the preset safety threshold.

[0119] Step S221: In response to the predicted value of dissolved gas concentration exceeding the safety threshold, an early warning is triggered and a status quantification report is generated. The status quantification report includes at least one of the following: a bubble size distribution bar chart, a frequency domain attenuation curve, and the concentration exceeding the standard.

[0120] In step S220, the host computer obtains the predicted dissolved gas concentration value output in step S200 and acquires a preset safety threshold. The preset safety threshold is set according to the dissolved gas concentration warning values ​​in oil specified in power industry standards, or according to the operating history data and maintenance records of the monitored equipment. For example, the preset safety threshold is set according to the dissolved gas concentration warning values ​​in oil specified in power industry standards, but it can also be dynamically adjusted based on the operating years of the monitored equipment, historical monitoring data, and equipment type. For instance, in the initial stage of equipment commissioning, the safety threshold is set to the warning value specified in power industry standards; as the operating time increases, the host computer obtains the predicted dissolved gas concentration values ​​at multiple time points under normal operating conditions, calculates their mean and standard deviation, and corrects the safety threshold to the mean plus three times the standard deviation, so that the safety threshold matches the operating characteristics of the equipment itself. The safety threshold is stored in the host computer for use in step S220.

[0121] The host computer compares the predicted dissolved gas concentration with the safety threshold to determine whether the predicted dissolved gas concentration exceeds the safety threshold.

[0122] In step S221, the early warning includes audible and visual alarm signals or remotely pushed alarm information. The status quantification report includes at least one of the following: a bubble size distribution histogram, a frequency domain attenuation curve, and the concentration exceedance magnitude. The bubble size distribution histogram is used to show the distribution of the number of bubbles within each radius interval; the frequency domain attenuation curve is used to show the trend of amplitude attenuation rate changes at each target excitation frequency; and the concentration exceedance magnitude is used to represent the relative magnitude by which the predicted dissolved gas concentration exceeds the safety threshold. It is calculated by subtracting the safety threshold from the predicted dissolved gas concentration and then dividing by the safety threshold to obtain the percentage exceedance magnitude. The status quantification report presents the detection results in a visual manner, providing maintenance personnel with intuitive equipment status information and a basis for maintenance decisions.

[0123] For details, please refer to Figure 8 Step S221 may include: Step S222: Extract the peak coordinates of the bubble size distribution histogram and the slope values ​​of the frequency domain attenuation curve.

[0124] Step S223: Calculate the degradation membership degree based on the peak coordinates and slope values.

[0125] Step S224: Determine the degradation level of the insulating oil based on the degradation membership degree.

[0126] In step S222, the state quantization report includes a bubble size distribution histogram and a frequency domain attenuation curve. The host computer performs peak detection on the bubble size distribution histogram, identifies the radius interval corresponding to the bar with the highest amplitude in the histogram, and uses the median of this radius interval as the peak coordinate. The host computer performs linear fitting on the frequency domain attenuation curve, calculates the slope of the fitted line, and obtains the slope value of the frequency domain attenuation curve.

[0127] In step S223, optionally, a pre-established degradation judgment matrix can be obtained. This matrix stores the membership functions of different degradation levels corresponding to combinations of peak coordinates and slope values. The host computer substitutes the peak coordinates and slope values ​​into the degradation judgment matrix and calculates the degree to which the insulating oil belongs to each degradation level through fuzzy logic reasoning, thus obtaining the degradation membership degree. The degradation membership degree is a value between 0 and 1, used to represent the degree to which the insulating oil belongs to each degradation level.

[0128] In step S224, the host computer compares the degradation membership degree with the membership degree threshold of each degradation level, and selects the degradation level with the highest degradation membership degree that exceeds the corresponding membership degree threshold as the current degradation level of the insulating oil. For example, when the degradation membership degree is 0.85, the degradation level of the insulating oil is determined to be Level 2, moderate degradation. The host computer outputs and stores the degradation level determination result for equipment condition assessment and maintenance decisions.

[0129] In this embodiment, the predicted dissolved gas concentration is first compared with a preset safety threshold to automatically determine the oil deterioration state. The safety threshold is set according to power industry standards and can be dynamically adjusted based on equipment operating history, improving the targeting and accuracy of the determination. In response to the predicted value exceeding the safety threshold, an early warning is triggered and a status quantification report is generated. The report includes at least one of the following: a bubble size distribution histogram, a frequency domain attenuation curve, and the concentration exceeding the standard. The detection results are presented visually, providing maintenance personnel with intuitive equipment status information and maintenance decision-making basis. Existing technologies typically only output concentration values, lacking visualization reports and early warning mechanisms. This embodiment achieves a complete decision-making chain from concentration prediction to early warning triggering through the above methods, improving the readability of the detection results and the decision support capability.

[0130] It is understandable that after integrating the above methods, the intelligent sensor can complete the entire process of ultrasonic echo acquisition, frequency domain transformation, anomaly identification, and concentration prediction in situ through its built-in ultrasonic transceiver unit and signal processing unit. This eliminates the need for offline sampling and laboratory testing, and allows for real-time output of predicted values ​​for trace moisture, particle size, acid value, and dissolved gas concentration. This enables comprehensive multi-parameter evaluation and early warning of insulating oil quality. The status quantification report presents the test results in a visual manner, providing maintenance personnel with intuitive equipment status information and a basis for maintenance decisions.

[0131] In one embodiment, see Figure 9 Step S160 includes: Step S166: Obtain the signal strength of the secondary radiation echo signal.

[0132] Step S167: Determine whether the signal strength is less than the preset strength threshold.

[0133] Step S168: In response to a signal strength greater than or equal to a preset strength threshold, the secondary radiation echo signal is transformed to the frequency domain.

[0134] In step S166, the host computer reads the digital signal of the secondary radiation echo signal acquired and stored in step S140, and calculates the root mean square value or peak amplitude of the signal as the signal strength, which reflects the overall energy level of the echo signal.

[0135] In step S167, the preset intensity threshold is the lowest acceptable signal intensity value statistically set after detecting multiple normal oil samples. The setting method is as follows: multiple known normal insulating oil samples are collected, and detection is performed according to steps S100 to S140. The secondary radiation echo signal intensity of each normal sample is calculated, and the lower quartile of the signal intensity distribution of all normal samples is taken as the preset intensity threshold. The host computer compares the signal intensity obtained in step S166 with this preset intensity threshold.

[0136] In step S168, if the signal strength is greater than or equal to the preset strength threshold, it indicates that the echo signal has a sufficient signal-to-noise ratio and can be effectively analyzed in the frequency domain. The host computer then executes the frequency domain transformation processing from steps S161 to S165. If the signal strength is less than the preset strength threshold, it indicates that the echo signal is too weak and the signal-to-noise ratio is insufficient. Directly performing frequency domain analysis may cause the effective signal in the spectrum to be submerged by noise. The host computer increases the transmission power of the ultrasonic pulse train and then re-executes the frequency sweep transmission and echo reception in step S140 until the signal strength meets the conditions before performing the frequency domain transformation.

[0137] In this embodiment, the signal strength of the secondary radiation echo signal is first acquired, and then compared with a preset intensity threshold. This intensity threshold is set after statistical analysis of multiple normal oil samples. When the signal strength is less than the threshold, the transmission power is increased and the signal is retransmitted until the signal strength meets the threshold before performing frequency domain transformation, thus avoiding the failure of frequency domain analysis results due to weak signals. Existing technologies lack signal quality judgment and adaptive adjustment mechanisms; when the signal strength is insufficient, the effective signal in the spectrum is overwhelmed by noise. This embodiment ensures the quality of the input signal for frequency domain analysis through the above methods, improving the stability and reliability of the detection results.

[0138] For the same purpose, in one embodiment, please refer to Figure 10 A multi-frequency ultrasonic-based on-load tap changer insulating oil attenuation detection system is provided, including: The bubble detection module is used to obtain the initial radius dataset of multiple bubbles in the insulating oil.

[0139] The frequency generation module is used to substitute the initial radius dataset into the Minsner equation to obtain the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation is used to represent the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence is used to represent the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship is used to represent the bubble radius corresponding to each resonant frequency.

[0140] The ultrasonic transceiver module is used to transmit ultrasonic pulse trains to the insulating oil according to the target excitation frequency sequence and to receive secondary radiation echo signals.

[0141] The frequency domain transformation module is used to transform the secondary radiation echo signal to the frequency domain to obtain resonant frequency domain response data. The resonant frequency domain response data is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency.

[0142] The abnormal accumulation module is used to obtain the cumulative value of abnormal energy loss based on the resonance frequency domain response data and the bubble mapping relationship. The cumulative value of abnormal energy loss is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance.

[0143] The concentration prediction module is used to obtain the predicted value of dissolved gas concentration based on the cumulative value of abnormal energy loss and the support vector machine regression model. The predicted value of dissolved gas concentration is used to represent the degree of deterioration of the insulating oil.

[0144] The bubble detection module, frequency generation module, ultrasonic transceiver module, frequency domain transformation module, anomaly accumulation module, and concentration prediction module in the host computer are used to execute the functions described in steps S100 to S200 and their respective sub-steps. All of these modules are deployed in the host computer and are integrated software functional units that can be deployed on a local server or a cloud computing platform. The modules are connected sequentially, with the output of the previous module connected to the input of the next module, forming a data transfer channel.

[0145] This application also provides an electronic device, which includes one or more processors and memory resources represented by memory for storing instructions executable by the processor, such as application programs. The application programs stored in the memory may include one or more modules, each corresponding to a set of instructions.

[0146] The electronic device may also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on operating devices stored in memory, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0147] In one embodiment, a computer device, which may be a server, is also provided. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device stores data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection.

[0148] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0149] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of the electronic device, enables the electronic device to perform one of the above embodiments.

[0150] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0151] It should be noted that although the steps of the method in this application are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps, such as omitting certain steps, combining multiple steps into one step, and / or breaking down one step into multiple steps, should all be considered part of this application.

[0152] It should be understood that this application is not limited to the detailed data of the methods presented in this specification. This application is capable of other implementations and can be implemented and performed in various ways. The foregoing variations and modifications fall within the scope of this application. It should be understood that the disclosure and definition of this application extend to all alternative combinations of two or more individual features mentioned or apparent in the text and / or drawings. All these different combinations constitute multiple alternative aspects of this application. The embodiments described in this specification illustrate the best known mode for implementing this application and will enable those skilled in the art to utilize this application.

Claims

1. A method for detecting the attenuation of insulating oil in on-load tap changers based on multi-frequency ultrasound, characterized in that, include: Obtain the initial radius dataset of multiple air bubbles in the insulating oil; Substituting the initial radius dataset into the Minsner equation yields the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation represents the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence represents the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship represents the bubble radius corresponding to each resonant frequency. The ultrasonic pulse train is transmitted to the insulating oil according to the target excitation frequency sequence, and the secondary radiation echo signal is received. The secondary radiation echo signal is transformed to the frequency domain to obtain resonant frequency domain response data, which is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency. Based on the resonant frequency domain response data and the bubble mapping relationship, the cumulative value of abnormal energy loss is obtained, which is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance. Based on the cumulative value of abnormal energy loss and the support vector machine regression model, a predicted value of dissolved gas concentration is obtained, which is used to represent the degree of deterioration of the insulating oil.

2. The method according to claim 1, characterized in that, The process of obtaining the initial radius dataset of multiple air bubbles in the insulating oil includes: The ultrasonic echo signal of the insulating oil in a resting state was collected; Perform a Hilbert transform on the echo signal to obtain the envelope curve; The backscattering sound pressure of each bubble is determined based on the peak position of the envelope curve. The acoustic scattering cross-sectional area of ​​each bubble is calculated based on the backscattering sound pressure of each bubble. The acoustic scattering cross-sectional area of ​​each bubble is regressed to obtain the initial radius value of each bubble; The initial radius dataset is generated based on the multiple initial radius values.

3. The method according to claim 2, characterized in that, The step of transforming the secondary radiation echo signal to the frequency domain to obtain resonant frequency domain response data includes: A windowed signal is obtained by applying a Hanning window function to the secondary radiation echo signal; Perform a Fast Fourier Transform on the windowed signal to obtain the frequency domain amplitude spectrum; Extract the peak amplitude at each target excitation frequency from the frequency domain amplitude spectrum; Based on each of the aforementioned peak amplitudes, determine the amplitude attenuation rate at each target excitation frequency; The resonant frequency domain response data is composed of each target excitation frequency and the corresponding amplitude attenuation rate.

4. The method according to claim 3, characterized in that, The step of obtaining the cumulative value of abnormal energy loss based on the resonant frequency domain response data and the bubble mapping relationship includes: Calculate the ratio of the amplitude attenuation rate to the initial emitted sound pressure at each target excitation frequency; Target excitation frequencies whose ratio exceeds a preset attenuation ratio threshold are marked as abnormal frequencies; The bubble radius range corresponding to the abnormal frequency is located by using the bubble mapping relationship. The cumulative value of abnormal energy loss is obtained by summing the acoustic scattering cross-sectional areas of all bubbles within the bubble radius range corresponding to the abnormal frequency.

5. The method according to claim 1, characterized in that, The step of transmitting an ultrasonic pulse train to the insulating oil according to the target excitation frequency sequence includes: The difference between adjacent frequencies in the target excitation frequency sequence is set as the sweep step interval; According to the frequency sweep step interval, ultrasonic pulse trains of each target excitation frequency are sequentially output to the insulating oil.

6. The method according to claim 1, characterized in that, The step of transmitting an ultrasonic pulse train to the insulating oil according to the target excitation frequency sequence further includes: Obtain the frequency response compensation coefficient at each target excitation frequency; A sound pressure amplitude modulation curve is generated based on the frequency response compensation coefficient; The sound pressure amplitude of the ultrasonic pulse train at each target excitation frequency is controlled according to the sound pressure amplitude modulation curve to keep the sound pressure amplitude constant at each target excitation frequency.

7. The method according to claim 1, characterized in that, After obtaining the predicted value of dissolved gas concentration based on the cumulative value of abnormal energy loss and the support vector machine regression model, the process further includes: The predicted value of the dissolved gas concentration is compared with a preset safety threshold. In response to the predicted value of dissolved gas concentration exceeding the safety threshold, an early warning is triggered and a status quantification report is generated. The status quantification report includes at least one of the following: a bubble size distribution bar chart, a frequency domain attenuation curve, and the concentration exceeding the standard.

8. The method according to claim 7, characterized in that, After generating the state quantification report, the following is also included: Extract the peak coordinates of the bubble size distribution histogram and the slope values ​​of the frequency domain attenuation curve; Calculate the degradation membership degree based on the peak coordinates and the slope value; The degradation level of the insulating oil is determined based on the degradation membership degree.

9. The method according to claim 1, characterized in that, The step of transforming the secondary radiation echo signal to the frequency domain further includes: Obtain the signal strength of the secondary radiation echo signal; Determine whether the signal strength is less than a preset strength threshold; In response to the signal strength being greater than or equal to a preset strength threshold, the secondary radiation echo signal is transformed to the frequency domain.

10. A multi-frequency ultrasonic-based on-load tap changer insulating oil attenuation detection system, characterized in that, include: The bubble detection module is used to obtain the initial radius dataset of multiple bubbles in the insulating oil. The frequency generation module is used to substitute the initial radius dataset into the Minsner equation to obtain the target excitation frequency sequence and the bubble mapping relationship. The Minsner equation is used to represent the physical relationship between the bubble radius and the natural resonant frequency. The target excitation frequency sequence is used to represent the frequency sweep range required to excite each bubble to resonate. The bubble mapping relationship is used to represent the bubble radius corresponding to each resonant frequency. An ultrasonic transceiver module is used to transmit ultrasonic pulse trains to the insulating oil according to the target excitation frequency sequence and to receive secondary radiation echo signals. The frequency domain transformation module is used to transform the secondary radiation echo signal to the frequency domain to obtain resonant frequency domain response data. The resonant frequency domain response data is used to represent the degree of absorption of ultrasonic energy by the insulating oil at each resonant frequency. An abnormal accumulation module is used to obtain an abnormal energy loss accumulation value based on the resonance frequency domain response data and the bubble mapping relationship. The abnormal energy loss accumulation value is used to represent the total amount of ultrasonic energy lost due to abnormal bubble resonance. The concentration prediction module is used to obtain a predicted value of dissolved gas concentration based on the cumulative value of abnormal energy loss and the support vector machine regression model. The predicted value of dissolved gas concentration is used to represent the degree of deterioration of the insulating oil.

Images