System and method for detecting free carbon particles in transformer oil based on microelectrode array

Abstract

The invention provides a system and method for detecting free carbon particles in transformer oil based on a microelectrode array, and the system comprises a sampling module which comprises an oil inlet pipeline and a filtering device, an oil sample is conveyed to a microelectrode array sensor module from a transformer oil tank through the oil inlet pipeline, and the filtering device is located in the oil inlet pipeline; a filter membrane is arranged in the filter device to remove impurities in the oil sample; the microelectrode array sensor module comprises a microelectrode array sensor, the microelectrode array sensor is nanoelectrodes arranged on a silicon substrate, the surface of the microelectrode array sensor is coated with graphene oxide to adsorb free carbon particles in an oil sample, and when the free carbon particles are adsorbed to the surface of the microelectrode array sensor, an adsorption signal is generated; and the signal processing unit is used for extracting the characteristic frequency and intensity of the vibration of the carbon particles in the adsorption signal, and classifying the carbon particles according to the characteristic frequency and intensity of the carbon particles. The device is compact in structure, the maintenance cost is reduced, and the operation process is simplified.

Description

Technical Field

[0001] The present invention belongs to the field of transformer impurity identification, and particularly relates to a free carbon particle detection system and method in transformer oil based on a microelectrode array. Background Art

[0002] As a core device for power transmission and distribution, the reliability of a transformer directly affects the safe operation of the power grid. Insulating oil, as an important medium inside the transformer, has the functions of insulation and cooling. However, with the long-term operation of the transformer, various impurities will be generated in the insulating oil, especially free carbon particles. These carbon particles are generated due to the deterioration of insulating materials, the decomposition of oil, etc. Their presence not only reduces the insulation performance of the oil but also may cause serious problems such as discharge and local overheating, and even lead to equipment failures. Therefore, the detection and monitoring technology for free carbon particles in transformer oil is particularly important. In the prior art, optical, filtration, and chemical analysis methods are mainly used to detect carbon particles, but these methods usually have problems such as low detection sensitivity, complex operation, and poor real-time performance, and it is difficult to meet the accurate monitoring requirements of the operation status of transformers in modern power systems.

[0003] The existing methods for detecting impurity particles in transformer oil mainly rely on offline detection technologies, which usually require multiple steps such as sampling, transportation, and laboratory analysis. The process is cumbersome and time-consuming. Since it is an offline detection, it is difficult to timely reflect the changes in impurity particles in transformer oil and cannot meet the requirements for real-time monitoring of the operation status of transformers. In addition, the existing detection equipment is usually large in volume, expensive in price, and high in maintenance cost, and is not suitable for on-site application. And it requires professional personnel to operate, increasing the labor cost and operation risk. Due to the existence of factors such as inability to perform real-time online detection, complex and time-consuming detection steps, high maintenance cost, and complex operation, the prior art is difficult to meet the requirements for efficient and accurate monitoring of impurity particles in transformer oil.

[0004] The prior art document 1 (CN 117213864 A, publication date: December 12, 2023) proposes an intelligent detection method and system for metal debris particles in the lubricating oil of an engine, including: obtaining a primary amplified signal, a secondary amplified signal, and a vibration signal corresponding to a metal debris particle sensor in the lubricating oil; performing noise reduction processing; and obtaining a metal debris particle identification result and a particle statistical distribution sequence in the observed oil path. The oil particle detection method provided by the prior art document 1 cannot adjust the oil sample flow rate in real time according to the concentration state of metal debris particles in the oil, and the detection error is relatively large.

[0005] The prior art document 2 (CN 116202915 A, publication date: June 2, 2023) discloses a method for testing impurity particles in lubricating oil by ferrography. The equipment oil sample is collected and injected into an oil sample bottle, heated in an oven at 65°C until the water content of the oil sample ≤ 20 ppm, then taken out and ultrasonically vibrated. Next, the iron element content of the oil sample is measured, and the oil sample is diluted according to its content. Then, the kinematic viscosity of the oil sample is measured, and the viscosity of the oil sample is diluted according to the kinematic viscosity by designing a ratio. Then, an analytical ferrograph is used to make a spectrum, and finally, a ferrographic microscope is used to observe the particles on the spectrum plate. First, observe with a low-power eyepiece. After the eyepiece field of view is clear, align the lens of the digital device with the image in the eyepiece, and fine-tune to make the image on the digital device clear, and then switch to a high-power lens for observation. The method for detecting particles in oil provided by the prior art document 2 requires laboratory analysis and cannot meet the requirements for real-time monitoring of the operating status of transformers. Summary of the Invention

[0006] To solve the deficiencies in the prior art, the present invention provides an on-line detection device for free carbon particles in transformer oil based on a microelectrode array sensor, which can realize real-time detection of free carbon particles in transformer oil, has the advantages of high detection accuracy, simple operation and low maintenance cost, can not only detect free carbon particles in oil in real time, but also identify changes in particle concentration, and has broad application prospects and market value.

[0007] The present invention adopts the following technical solutions.

[0008] The present invention provides an on-line detection system for free carbon particles in transformer oil based on a microelectrode array, including: a sampling module, a microelectrode array sensor module, a signal processing unit, a communication module, a remote monitoring platform and an alarm unit;

[0009] The sampling module includes an oil inlet pipeline and a filtering device. The oil sample is transported from the transformer oil tank to the microelectrode array sensor module through the oil inlet pipeline. The filtering device is located inside the oil inlet pipeline, and a filter membrane is arranged inside the filtering device to remove impurities in the oil sample;

[0010] The microelectrode array sensor module includes a microelectrode array sensor. The microelectrode array sensor is a nanoelectrode arranged on a silicon substrate. The surface of the microelectrode array sensor is coated with graphene oxide to adsorb free carbon particles in the oil sample. When the free carbon particles are adsorbed on the surface of the microelectrode array sensor, an adsorption signal is generated;

[0011] The signal processing unit is used to extract the characteristic frequency and intensity of the vibration of carbon particles in the adsorption signal, and classify the carbon particles according to the characteristic frequency and intensity of the carbon particles.

[0012] Preferably, the sampling module further includes a servo motor, a micro pump, and a valve; the valve, the micro pump, and the servo motor are all located outside the electrostatic shielding layer, and the electrostatic shielding layer is coated outside the oil inlet pipe; the micro pump is used to push the oil sample to flow, the valve is used to adjust the flow rate of the oil sample, and the servo motor is used to control the opening degree of the valve to adjust the flow rate of the oil sample.

[0013] Preferably, the microelectrode array sensor module further includes a temperature sensor and a pressure sensor; the microelectrode array sensor, the temperature sensor, and the pressure sensor are all arranged in the oil product detection bin; the oil outlet pipe is connected to the oil product detection bin; the pressure sensor is connected to the servo motor through a servo motor control line.

[0014] Preferably, the microelectrode array sensor adopts a regular hexagonal honeycomb arrangement, and the nanoelectrodes are arranged at equal intervals on a square silicon substrate with a side length of c; from the first row of nanoelectrodes to the row with the largest number of nanoelectrodes, the distances from the leftmost and rightmost nanoelectrodes to the edge of the silicon substrate decrease row by row, and the decrease amount is Δd; the distances from the leftmost and rightmost nanoelectrodes in the row with the largest number of nanoelectrodes to the edge of the silicon substrate are d max ; from the row with the largest number of nanoelectrodes to the last row of nanoelectrodes, the distances from the leftmost and rightmost nanoelectrodes to the edge of the silicon substrate increase row by row, and the increase amount is Δd.

[0015] The present invention also provides an on-line detection method for free carbon particles in transformer oil based on a microelectrode array, which is realized by using the aforementioned on-line detection system for carbon particles in transformer oil based on a microelectrode array, and includes:

[0016] The oil sample obtained by the sampling module flows into the microelectrode array sensor module through the oil inlet pipe; the pressure and temperature of the oil sample are monitored in real time by the pressure sensor and the temperature sensor of the microelectrode array sensor module, and the micro pump and the valve of the sampling module are adjusted according to the real-time change values of the pressure and temperature of the oil sample to adjust the flow rate and flow velocity of the oil sample;

[0017] After the flow rate and flow velocity of the oil sample are adjusted, the microelectrode array sensor generates an adsorption signal when adsorbing free carbon particles in the oil sample;

[0018] The signal processing unit extracts the characteristic frequency and intensity of the particle vibration in the adsorption signal, and classifies the carbon particles according to the characteristic frequency and intensity of the carbon particles.

[0019] Preferably, the micro pump of the sampling module is adjusted according to the real-time change values of the pressure and temperature of the oil sample, and the flow rate of the micro pump after adjustment satisfies the following relational formula:

[0020]

[0021] In the formula, Q adjust is the flow rate of the micro pump after adjustment, and the unit is m3 / s; K p is the first proportionality coefficient, with a value of 1.2; P target is the target pressure, in MPa; P actual is the actually collected oil sample pressure, in MPa; μ(T) is the viscosity of the oil sample, in Pa·s, determined by the temperature T, with a viscosity influence coefficient α of 0.8; ΔP is the pressure change of the oil sample, in MPa, with an influence coefficient β of 0.5; C part is the carbon particle concentration in the oil sample, in mg / L; C0 is the initial carbon particle concentration, in mg / L; γ and δ are the linear and non-linear influence coefficients of the particle concentration on the flow rate adjustment respectively; D particle is the particle size of the carbon particles, in μm; K viscosity is the coefficient related to viscosity and particle size; λ1 is the adjustment coefficient of the particle size on the viscosity; ζ is the non-linear influence coefficient of the particle concentration on the flow rate adjustment.

[0022] Preferably, the valve of the sampling module is adjusted according to the real-time change values of the pressure and temperature of the oil sample, and the adjusted valve opening satisfies the following relational expression:

[0023]

[0024] where, α adjust is the adjusted valve opening, in °, α0 is the initial valve opening, in °, γ1 is the influence coefficient of the temperature on the valve opening, with a value of 0.3, ΔT is the temperature change, in °C; γ2 is the influence coefficient of the pressure change on the valve opening, with a value of 0.4, ΔP is the pressure change of the oil sample, in MPa; C part is the carbon particle concentration in the oil sample, in mg / L, C0 is the initial carbon particle concentration, in mg / L, β1 is the influence coefficient of the particle concentration on the valve opening, with a value of 0.5, η is the non-linear response coefficient of the particle concentration, with a value of 1.2; Q adjust is the flow rate after the adjustment of the micro pump, in m 3 / s, Q max is the maximum flow rate, in m 3 / s, referring to the flow rate when only the valve opening is adjusted to the maximum; ζ v is the secondary influence coefficient of the flow rate on the valve opening; D particle is the particle size of the carbon particles, in μm, α2 is the influence coefficient of the particle size on the valve opening; γ3 is the response coefficient of the particle concentration on the valve opening, k is the adjustment coefficient of the Sigmoid function.

[0025] Preferably, adjusting the flow velocity and flow rate of the oil sample further includes:

[0026] Adjustment is carried out by combining PID control and machine learning algorithms, and the control output satisfies the following relational expression:

[0027]

[0028] where u(t) is the control output, the adjustment amount of the micro-pump flow rate or valve opening; e(t) is the error signal, the difference between the target pressure and the actual pressure, and t is the oil sample detection duration; K c is the second proportionality coefficient, and its value is 0.8; K i is the integral coefficient, and its value is 0.5; K d is the differential coefficient, and its value is 0.3; ω1 is the adjustment coefficient of the flow rate to the PID feedback; Q actual is the actual flow rate, unit: m 3 / s, Q max is the maximum flow rate, unit: m 3 / s; ω2 is the adjustment coefficient of the particle concentration to the PID feedback; C part is the carbon particle concentration in the oil sample, unit: mg / L, C0 is the initial carbon particle concentration, unit: mg / L; μ is the non-linear response coefficient of the particle concentration to the detection sensitivity.

[0029] Preferably, the signal processing unit analyzes and processes the adsorption signal using the Fourier transform algorithm and the wavelet transform algorithm, converts the time-domain signal detected by the microelectrode array sensor into a frequency-domain signal, extracts the characteristic frequency and intensity of the free carbon particles, the characteristic frequency includes the resonance frequency offset of the carbon particles, and the intensity includes the resonance signal attenuation intensity of the carbon particles. The wavelet transform algorithm is used to distinguish carbon particles of different particle sizes, and the SVM algorithm is used to classify the data to judge the content of carbon particles in the oil sample.

[0030] Preferably, the signal processing unit classifies the carbon particles according to the characteristic frequency and intensity of the carbon particles, including:

[0031] Classify according to the resonance frequency offset Δf of the adsorption signal and the resonance signal attenuation intensity I. When 0 < Δf ≤ 10 kHz and I ≤ 20 dB, the carbon particles are determined to be low-concentration small particles; when 0 < Δf ≤ 10 kHz and I > 20 dB, the carbon particles are determined to be high-concentration small particles; when Δf > 10 kHz, the carbon particles are determined to be large-size conductive particles.

[0032] The beneficial effect of the present invention is that, compared with the prior art,

[0033] 1. The present invention adopts a microelectrode array sensor, which can efficiently detect free carbon particles in transformer oil by coating a nanoelectrode array with graphene oxide. Graphene oxide has excellent conductivity and a large specific surface area, which enhances the sensor's adsorption and reaction efficiency to carbon particles. The sensor monitors the adsorption of carbon particles and current changes through electrochemical reactions, has high sensitivity, and can detect low concentrations of carbon particles. The high surface area and fine structure of the nanoelectrode ensure high-resolution detection performance, can accurately capture small signal changes, and further improve detection accuracy and stability through advanced signal amplification and processing technology, providing real-time and accurate carbon particle concentration data. A layer of nanocoating based on bionics principles is provided on the surface of the nanoelectrode. By constructing a nanoscale rough structure and low surface energy materials, superoleophobic properties are achieved, thereby preventing carbon particles from depositing on the electrode surface.

[0034] 2. During the oil sample collection process, the present invention comprehensively considers the oil sample pressure, viscosity, particle concentration and particle size, and dynamically adjusts the oil sample flow rate and flow rate by adjusting the micro pump and valve opening, thereby ensuring the flow stability of the oil sample under different conditions, avoiding particle deposition or uneven flow, and especially maintaining high precision in the detection of small-size carbon particles.

[0035] 3. The signal processing unit in the present invention integrates advanced signal processing algorithms such as Fourier transform and wavelet transform, which can efficiently process the detection signal, filter out noise and extract key signal features, thereby ensuring the accuracy of the detection results.

[0036] 4. The device of the present invention has a compact structure and is easy to install. It is suitable for on-site online detection of transformers, reduces maintenance costs, and simplifies the operation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of an online detection system for carbon particles in transformer oil based on a microelectrode array in the present invention;

[0038] Figure 2 Schematic diagram of the microelectrode array in the present invention;

[0039] Figure 3 It is a schematic diagram of the working principle of the sampling module and the microelectrode array sensor module in the present invention;

[0040] Figure 4 It is a schematic diagram of the signal processing unit and the communication system in the present invention;

[0041] Among them, the main marks of the drawings in the figure are:

[0042] 1 - Silicon substrate; 2 - Nanoelectrode, 21 - Papillary structure; 3 - Filter device; 31 - Filter membrane; 4 - Oil inlet pipe; 5 - Oil outlet pipe; 6 - Oil quality detection chamber; 7 - Microelectrode array sensor; 8 - Electrostatic shielding layer; 9 - Pressure sensor; 10 - Valve; 11 - Servo motor; 12 - Servo motor control line; 13 - Pressure signal processor; 14 - Micro pump. Detailed implementation manners

[0043] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described in this application are only a part of the embodiments of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

[0044] An on - line detection system for carbon particles in transformer oil based on a microelectrode array provided by the present invention is characterized in that the device includes: a sampling module, a microelectrode array sensor module, a signal processing unit, a communication module, and an alarm unit.

[0045] The sampling module includes an oil inlet pipe 4, a valve 10, a micro pump 14, a servo motor 11, a servo motor control line 12, and a filter device 3. The outer layer of the oil inlet pipe 4 is coated with an electrostatic shielding layer 8; the valve 10, the micro pump 14, and the servo motor 11 are located outside the electrostatic shielding layer 8. The servo motor 11 is connected to the pressure signal processor 13 through the servo motor control line 12. The servo motor 11 is used to control the valve 10 to adjust the oil sample flow rate; the filter device 3 is located inside the oil inlet pipe 4, and the filter membrane 31 is located inside the filter device 3 for removing impurities in the oil sample.

[0046] The micro pump 14 is used to provide stable flow, with a flow range of 0.1 - 1 L / min, to ensure that the oil sample can be evenly transmitted to the subsequent detection module. The valve 10 plays a role in adjusting the flow rate of the oil sample, usually controlling the flow rate of the oil sample between 1 - 5 mL / min to ensure the stability of the oil sample flow rate and avoid particle deposition or uneven flow. The micro pump 14 and the valve 10 each perform different functions and work together to ensure the stable transmission of the oil sample to the sensor module for detection.

[0047] The oil sample is filtered by the filter device 3 before entering the detection device to ensure the detection accuracy. To avoid carbon particles adhering to the sampling pipeline due to electrostatic action during the sampling process, the system integrates an electrostatic shielding layer 8 to ensure that the suspended state of carbon particles during flow remains unchanged. The sampling module controls the flow rate of the oil sample through the micro pump 14 and the valve 10, and the flow rate is controlled between 1 and 5 milliliters per minute to ensure the stability of the oil sample flow rate and avoid particle deposition or uneven flow. The system also monitors the temperature and pressure of the oil sample in real time. The temperature range is set from 40°C to 80°C, and when it exceeds 85°C, the cooling device is activated; the pressure range is set from 0.5 to 2 MPa. If the pressure fluctuation is greater than ±0.1 MPa, the system automatically adjusts the micro pump 14 and the valve 10 to ensure that the temperature and pressure of the oil sample are in a stable state and avoid affecting the properties and distribution of particles.

[0048] In the present invention, in order to improve the detection accuracy of small-sized carbon particles, multiple improvements have been made to the flow rate adjustment algorithm of the micro pump 14. Small-sized carbon particles are easily affected by the flow rate. Too fast a flow rate may cause particle aggregation, and too slow a flow rate may cause particle deposition. Therefore, precise flow rate adjustment is crucial for the detection of small-sized carbon particles. Particles with a smaller particle size usually tend to deposit at low flow rates and aggregate into clusters at high flow rates. To prevent this phenomenon from occurring, the flow rate adjustment algorithm adds an influence term of the particle size on the flow resistance. In addition, the present invention provides an improved flow rate adjustment algorithm for the micro pump 14. By comprehensively considering factors such as the pressure, viscosity, particle concentration, and particle size of the oil sample, the flow rate of the oil sample is dynamically adjusted to improve the accuracy and stability of particle detection in the microelectrode array in oil. The formula of the flow rate adjustment algorithm is:

[0049]

[0050] Where: Q adjust is the flow rate after adjustment of the micro pump 14, with the unit of m 3 / s; K p is the first proportionality coefficient, and its value is calibrated to be 1.2 through experiments; P target is the target pressure, with the unit of MPa; P actual is the actually collected oil sample pressure, with the unit of MPa; μ(T) is the viscosity of the oil sample, with the unit of Pa·s, which is determined by the temperature T. Its viscosity influence coefficient α is calibrated to be 0.8 through experiments; ΔP is the pressure change amount of the oil sample. The pressure change amount of the oil sample is the difference between the actually collected oil sample pressure and the initial oil sample pressure, with the unit of MPa. Its influence coefficient β is calibrated to be 0.5 through experiments; G part is the carbon particle concentration in the oil sample, with the unit of mg / L; C0 is the initial carbon particle concentration, with the unit of mg / L; γ and δ are respectively the linear influence coefficient and the non-linear influence coefficient of the particle concentration on the flow rate adjustment, and their values are calibrated to be 1.0 and 1.5 respectively through experiments; Dparticle is the particle size of carbon particles, with the unit of μm, and its particle size influence coefficient δ is calibrated to be 0.6 through experiments; K viscosity is the coefficient related to viscosity and particle size, and its value is calibrated to be 0.9 through experiments; λ1 is the adjustment coefficient of the influence of particle size on viscosity, and its value is calibrated to be 0.4 through experiments; ζ is the non-linear influence coefficient of particle concentration on flow rate adjustment, and its specific value is calibrated through experiments, and its value is calibrated to be 0.075 through experiments.

[0051] Target pressure P target is the ideal oil sample pressure preset by the system, and the actual pressure P actual is collected in real time through the pressure sensor 9. (P target -P actual ) in the formula represents the difference between the target pressure and the actual pressure, reflecting the deviation of the oil sample flow state. The system adjusts the flow rate of the micro pump 14 according to this pressure difference. The first proportional coefficient K p regulates the influence degree of the pressure difference on the flow rate. The viscosity μ(T) will change with the temperature. The viscosity of the oil sample decreases with the increase of temperature, and the fluidity increases, which in turn affects the flow rate. The influence of viscosity on the flow rate is regulated by the coefficient α to ensure the stable flow of the oil sample at different temperatures. The pressure change amount ΔP reflects the instability of the oil sample flow, and the influence of pressure fluctuation on the flow rate adjustment is regulated by the coefficient β.

[0052] The carbon particle concentration C in the oil sample part increase will increase the viscosity of the fluid, which in turn affects the flow rate. The influence of particle concentration on the flow rate is non-linear, and an exponential relationship is used for modeling. The coefficients γ and δ respectively regulate the linear and non-linear influences of concentration on the flow rate adjustment. For small particle size carbon particles, their fluidity is more easily affected by the change of flow velocity. D particle in the formula takes into account the non-linear influence of particle size on the flow rate. Smaller particles are easily affected by the change of flow velocity. Too high or too low flow velocity will affect the particle distribution and detection accuracy. By adjusting δ, the system can optimize the flow velocity to ensure that small particles are not affected by too high flow velocity and agglomerate.

[0053] First, the system calculates the amplitude of flow rate adjustment according to the difference between the target pressure and the actual pressure. The first proportional coefficient K p represents the response degree of flow rate adjustment to the pressure difference. When the difference between the target pressure and the actual pressure is large, the system will adjust the flow rate through the micro pump 14 to ensure that the oil sample flow velocity is consistent with the target flow velocity. Next, the flow rate adjustment also takes into account the influence of the temperature of the oil sample on the fluidity. Through μ(T) in the formula, the system can consider the influence of temperature change on the flow rate in real time, so as to avoid the flow velocity instability caused by temperature fluctuation.

[0054] In addition, the carbon particle concentration in the oil sample has a direct impact on fluidity. An increase in particle concentration increases the flow resistance of the oil sample, thereby affecting the stability of flow velocity and flow rate. For this reason, a non-linear influence term of particle concentration is introduced into the system. When the particle concentration is high, the system compensates for the increased flow resistance by reducing the flow rate. The influence of particle size is also taken into account, especially for carbon particles with small particle sizes, which are very sensitive to changes in flow velocity. By introducing particle size the system can avoid particle aggregation caused by excessive flow velocity and ensure the uniform distribution of small particles.

[0055] Therefore, multiple factors are considered in the entire flow rate adjustment process: pressure difference, temperature of the oil sample, particle concentration, particle size, etc. The system collects data such as pressure, temperature, particle concentration, and particle size in real time, substitutes these factors into the formula to calculate the flow rate adjustment amplitude, and controls the rotation speed of the micro pump 14 through a stepper motor to precisely adjust the flow rate. This adjustment ensures the flow stability of the oil sample under different conditions, avoids particle deposition or uneven flow, especially the detection of small particle size carbon particles, and maintains high precision.

[0056] To precisely control the flow velocity of the oil sample, the present invention proposes an improved valve opening adjustment formula, which comprehensively considers the influence of multiple factors such as temperature, pressure, particle concentration, flow rate, and particle size, and dynamically adjusts the opening of valve 10 to ensure the stability of the oil sample flow velocity. The formula is:

[0057]

[0058] Where: α adjust is the adjusted opening of valve 10 (unit: °), α0 is the initial opening of valve 10 (unit: °), γ1 is the influence coefficient of temperature on the opening of valve 10, and its value is calibrated to be 0.3 through experiments, ΔT is the temperature change amount (unit: °C). γ2 is the influence coefficient of pressure change on the opening of valve 10, and its value is calibrated to be 0.4 through experiments, ΔP is the pressure change amount of the oil sample (unit: MPa). C part is the carbon particle concentration in the oil sample (unit: mg / L), C0 is the initial concentration (unit: mg / L), β1 is the influence coefficient of particle concentration on the opening of valve 10, and its value is calibrated to be 0.5 through experiments, η is the non-linear response coefficient of particle concentration, and its value is calibrated to be 1.2 through experiments. Q adjust is the flow rate adjusted by the micro pump 14 (unit: m 3 / s), Q max is the maximum flow rate (unit: m 3 / s), which refers to the flow rate when only the opening of valve 10 is adjusted to the maximum; ζ vis the secondary influence coefficient of flow rate on the opening of valve 10, which is set to 0.5 for linear valves and 2.0 for non-linear valves. D particle is the particle size of carbon particles (unit: μm), α2 is the influence coefficient of particle size on the opening of valve 10, and its value is calibrated to be 0.6 through experiments. γ3 is the response coefficient of particle concentration on the opening of valve 10, and its value is calibrated to be 0.7 through experiments. k is the adjustment coefficient of the Sigmoid function, and its value is calibrated to be 0.9 through experiments.

[0059] By introducing the Sigmoid function into this formula the influence of particle concentration becomes smoother, avoiding violent fluctuations in the opening of valve 10 when the concentration changes too much. In addition, the formula also considers the secondary influence of flow rate on the opening of valve 10 By adjusting the relationship between flow rate and the opening of valve 10, the stability of the oil sample flow rate is ensured.

[0060] When implementing the adjustment of the opening of valve 10, the system first collects the temperature change ΔT, pressure fluctuation ΔP, particle concentration C part , flow rate Q adjust and particle size D particle of the oil sample in real time. The temperature change will affect the viscosity of the oil sample, and thus affect the flow rate. The system responds to the temperature change through γ1·ΔT and adjusts the opening of valve 10 in a timely manner to minimize the influence of temperature fluctuation on the flow rate. For example, when the temperature of the oil sample rises, the viscosity decreases, the flow rate increases, and the opening of valve 10 needs to be appropriately reduced. At the same time, the system will monitor the influence of the pressure change amount ΔP on the flow rate of the oil sample. When the pressure fluctuation is large, the opening of valve 10 will be appropriately increased to ensure stable flow and avoid uneven particle distribution.

[0061] The increase in particle concentration C part will lead to a decrease in the fluidity of the oil sample, and the opening of valve 10 needs to be reduced. Through the and the Sigmoid function term in the formula, the system can accurately adjust the opening of valve 10 to ensure that the influence of particle concentration change on the flow rate is effectively controlled. For example, when the particle concentration is high, the oil sample flow becomes more difficult, and the opening of valve 10 will be appropriately reduced to maintain a stable flow rate.

[0062] The change in flow rate Q adjust directly affects the flow rate of the oil sample. The system feeds back the change in flow rate through and adjusts the opening of valve 10. When the flow rate is too large, the flow rate of the oil sample increases, which may cause the particles to be unevenly distributed. At this time, the opening of valve 10 will be reduced to ensure an appropriate flow rate and avoid the particles flowing too fast or too slow.

[0063] In addition, the particle size D particleThe influence on the opening degree of valve 10 cannot be ignored. Small-sized carbon particles are particularly vulnerable to changes in flow velocity. Excessive flow velocity may cause particle aggregation, while too slow flow velocity will lead to particle deposition. The system adjusts the opening degree of valve 10 through α2·D in the formula particle to ensure the uniform distribution and stable flow of small particles.

[0064] During the implementation of the adjustment of the opening degree of valve 10, the system will collect data such as the flow rate, temperature, pressure, particle concentration, and particle size of the oil sample in real time, and substitute these data into the above formula for calculation. Then, the system precisely adjusts the opening degree of valve 10 through the servo motor 11 to ensure the stable flow rate of the oil sample and avoid particle deposition or aggregation. Especially when detecting small-sized carbon particles, high-precision detection is ensured.

[0065] To improve the response speed and detection accuracy of the system and address the influence of various factors such as particle concentration, flow rate, and pressure fluctuations on the system, the present invention combines PID control and machine learning optimization algorithms to further optimize the flow rate and the adjustment of the opening degree of valve 10. PID control adjusts the rotation of the servo motor 11 through real-time feedback to regulate the flow rate of the micro pump 14 and the opening degree of valve 10 to achieve the optimal system response:

[0066]

[0067] where: u(t) is the control output, representing the adjustment amount of the flow rate of the micro pump 14 or the opening degree of valve 10; e(t) is the error signal, that is, the difference between the target pressure and the actual pressure; K c is the second proportional coefficient, and its value is calibrated to be 0.8 through experiments; K i is the integral coefficient, and its value is calibrated to be 0.5 through experiments; K d is the differential coefficient, and its value is calibrated to be 0.3 through experiments; ω1 is the adjustment coefficient of the flow rate to the PID feedback, and its value is calibrated to be 0.7 through experiments; Q actual is the actual flow rate (unit: m 3 / s), Q max is the maximum flow rate (unit: m 3 / s); ω2 is the adjustment coefficient of the particle concentration to the PID feedback, and its value is calibrated to be 0.9 through experiments; C part is the carbon particle concentration in the oil sample (unit: mg / L), C0 is the initial concentration (unit: mg / L); μ is the non-linear response coefficient of the particle concentration to the detection sensitivity, and its value is calibrated to be 1.3 through experiments.

[0068] This formula realizes the precise adjustment of the flow rate of the micro pump 14 and the opening degree of valve 10 by combining the PID control principle and the feedback mechanism of the flow rate and particle concentration. During the operation of the system, first, the target pressure P target and the actual pressure Pactual , calculate the error signal e(t). The error signal e(t) is used to generate the proportional term, integral term, and derivative term, which are adjusted by coefficients K c , K i , and K d respectively. The proportional term is used to quickly respond to the current error, the integral term is used to eliminate the long-term steady-state error, and the derivative term is used to predict the trend of error change and reduce system overshoot and oscillation.

[0069] To further optimize the system response, the formula introduces feedback terms for flow rate and particle concentration. The flow rate feedback term is used to control the secondary influence of the flow rate on the system response and ensure that the system remains stable when the flow rate changes. The particle concentration feedback term is used to adjust the influence of the particle concentration on the system. Through the non-linear response coefficient μ, the change of the particle concentration has a smoother influence on the control quantity.

[0070] During the implementation process, the system substitutes the real-time collected data of the target pressure, actual pressure, flow rate, and particle concentration into the above formula to calculate the control quantity u(t), and precisely adjusts the flow rate of the micro pump 14 and the opening degree of the valve 10 through the servo motor 11. For example, when the difference between the target pressure and the actual pressure is large, the error signal e(t) increases, and the system quickly adjusts the control quantity through the proportional term to reduce the error. When the particle concentration increases, by dynamically adjusting the control quantity to ensure that the fluidity of the oil sample is not affected. In addition, the flow rate feedback term can appropriately reduce the control quantity when the flow rate is too large to avoid uneven particle distribution caused by too fast flow velocity.

[0071] As Figure 2 shown, the microelectrode array sensor 7 adopts a hexagonal honeycomb arrangement, and the nano-scale electrodes 2 are arranged on the silicon substrate 1 at equal intervals. Specifically, the diameter of the nano-electrode 2 is 50 nm, and the distance between the electrodes 2 is precisely controlled within the range of 50 - 100 nm, forming a high-density and low-interference detection network.

[0072] In this embodiment, the microelectrode array sensor 7 adopts a regular hexagonal honeycomb arrangement, and the nano-electrodes 2 are arranged on the square silicon substrate 1 with a side length of c at equal intervals; from the first row of nano-electrodes 2 to the row with the largest number of nano-electrodes 2, the distances from the leftmost and rightmost nano-electrodes 2 to the edge of the silicon substrate 1 decrease row by row, and the reduction amount is Δd, Δd = 50 mm; the distances from the leftmost and rightmost nano-electrodes 2 in the row with the largest number of nano-electrodes 2 to the edge of the silicon substrate 1 are d max = 150 mm; from the row with the largest number of nano-electrodes 2 to the last row of nano-electrodes 2, the distances from the leftmost and rightmost nano-electrodes 2 to the edge of the silicon substrate 1 increase row by row, and the increase amount is Δd; the row spacing of the nano-electrodes in the vertical direction is 86.6 nm;

[0073] A total of N nanoelectrodes 2 are arranged in the first row. Each nanoelectrode 2 has a diameter of 50 nm, and the center-to-center distance between the nanoelectrodes 2 is 100 nm. The distances from the leftmost and rightmost nanoelectrodes 2 in the first row of nanoelectrodes 2 to the edge of the silicon substrate 1 are both d1, where The second row arranges N + 1 nanoelectrodes 2; the third row arranges a total of N + 2 nanoelectrodes 2; it increases sequentially until the row with the largest number of nanoelectrodes 2, and then decreases sequentially until the nanoelectrodes 2 are arranged in a regular hexagon. This hexagonal honeycomb arrangement can not only maximize the utilization of the surface area of the silicon substrate 1 but also ensure the uniformity of the electric field distribution, significantly improving the capture efficiency and detection sensitivity of free carbon particles in transformer oil. Preferably, a nanoscale-thick graphene oxide material is coated on the sensor surface. Due to its unique two-dimensional structure and high specific surface area, graphene oxide can efficiently capture free carbon particles in the oil sample through π-π interaction and electrostatic adsorption mechanisms. In addition, the excellent conductivity of graphene oxide further enhances the response ability of the electrode to the change in conductivity after the adsorption of carbon particles. The material of the sensor electrode is preferably platinum, which has good conductivity, chemical stability, and corrosion resistance, and can work stably for a long time in the high-temperature and high-humidity environment of transformer oil, ensuring the reliability and service life of the detection device.

[0074] To further improve the performance of the microelectrode array, especially to prevent carbon particles from depositing on the electrode surface and enhance the long-term stability of detection, the embodiment of the present invention proposes a nano-coating design based on the principle of bionics. The bionic coating design is inspired by the microscopic structure of the lotus leaf surface, and its superhydrophobic property makes water droplets form spherical shapes on its surface and roll off easily. A similar principle can be applied to the microelectrode surface. By constructing a nanoscale rough structure and a low-surface-energy material, superoleophobic properties can be achieved, thereby preventing carbon particles from depositing on the electrode surface.

[0075] The surface microscopic structure of the coating adopts a nanoscale columnar or conical array, similar to the papilla structure 21 on the lotus leaf surface. This structure can significantly increase the surface roughness, thereby enhancing the oleophobic performance. The specific formula is as follows:

[0076] θ CB = arccos(r·cosθ0 + f - 1)

[0077] where, θ CB is the apparent contact angle (Cassie-Baxter model), r is the roughness factor, θ0 is the intrinsic contact angle of the material, and f is the liquid-solid contact area fraction. In order to achieve the apparent contact angle θ through the Cassie-Baxter model CBClose to or exceeding 150°, thus obtaining superoleophobic properties, systematic optimization of the roughness factor r and the contact area fraction f is required. By constructing a nanoscale columnar array to increase the surface roughness, the diameter of the columnar structure is 50 - 100 nm and the height is 200 - 500 nm, which can significantly increase the value of r. The surface material of the coating is selected as a low surface energy substance, such as polytetrafluoroethylene (PTFE) or fluoropolymer, whose surface energy is lower than 20 mJ / m 2 . This material can further reduce the adhesion force between carbon particles and the electrode surface, and can also significantly reduce the liquid-solid contact area fraction f. The oleophobic performance of the coating is evaluated by a contact angle measuring instrument and a rolling angle tester. The target value is the contact angle θ CB > 150 and the rolling angle < 5°. The specific formula is as follows:

[0078] cosθ CB =φ·cosθ0+(φ - 1)

[0079] where φ is the solid-liquid contact area fraction.

[0080] The liquid-solid contact area fraction f: Describes the proportion of the contact area between the liquid and the solid surface from the liquid perspective.

[0081] The solid-liquid contact area fraction φ: Describes the proportion of the contact area between the solid and the liquid surface from the solid perspective.

[0082] The liquid-solid contact area fraction f focuses more on the contact between the liquid and the solid, while the solid-liquid contact area fraction φ focuses more on the contact between the solid and the liquid. Strictly speaking, they are essentially the same in nature, both used to describe the proportion of the contact area between the liquid and the solid surface. In practical applications, f and φ can be used interchangeably, and the goal is to reduce the direct contact between the liquid and the solid surface by optimizing the surface microstructure and material properties, thereby enhancing the liquid repellent performance.

[0083] The liquid-solid contact area fraction f or the solid-liquid contact area fraction φ can be deduced inversely in combination with the Cassie-Baxter model. First, use a contact angle measuring instrument to measure the apparent contact angle θ of the coating CB , the intrinsic contact angle of polytetrafluoroethylene (PTFE) is about 110° - 120°, and the intrinsic contact angle of fluoropolymer is usually between 90° - 120°, depending on the surface treatment method of the materials produced by different manufacturers. On this basis, use the formula in the Cassie-Baxter model for calculation to obtain the specific values of the liquid-solid contact area fraction f or the solid-liquid contact area fraction φ. The specific formula is as follows:

[0084]

[0085] θ airis the contact angle between the liquid and air, with a value of 180°, so cosθ air = -1. θ Y is the intrinsic contact angle of the material, which is the contact angle of the material on an ideally smooth plane and is also known as the Young contact angle. θ CB is the apparent contact angle, which is the contact angle directly measured through experiments and represents the contact angle of the liquid on the actual surface of the coating (with a micro-rough structure).

[0086] In addition, in order to improve the mechanical stability and durability of the coating, a multi-layer composite structure design is adopted: an adhesion layer with good bonding properties to the electrode substrate material (such as silicon or platinum) at the bottom layer, such as silica or alumina. A functional layer with anti-corrosion and insulation properties in the middle layer, such as graphene oxide. A low surface energy oil-repellent layer is used on the surface layer, such as polytetrafluoroethylene (PTFE) or fluoropolymer. The signal processing unit integrates a microprocessor and an efficient signal processing algorithm, which can quickly process and analyze sensor data and accurately judge the carbon particle concentration in the oil sample. The communication module uses wireless transmission technology to send the processed data to the remote monitoring center in real time. The alarm unit is used to issue an alarm when the carbon particle concentration exceeds the standard to ensure that the operator can respond in time.

[0087] As Figure 4 shown, the signal processing unit uses the Fourier transform algorithm and the wavelet transform algorithm to analyze and process the electrical signals in the oil sample. The Fourier transform can convert the time-domain signal detected by the sensor into a frequency-domain signal and extract the characteristic frequencies related to free carbon particles. The specific formula is:

[0088]

[0089] In the formula, X(f) is the frequency-domain characteristic of carbon impurity particles in transformer oil. It describes the periodic changes caused by the vibration or other physical processes (such as flow, collision) of carbon particles in the oil, especially the response intensity at specific frequencies. x(t) is the signal that changes with time collected by the sensor, which represents the instantaneous signal detected by the sensor (such as an optical or electrical sensor) when carbon particles move, settle or distribute in the oil. f1 is the frequency, which represents the vibration frequency of carbon impurity particles in transformer oil or the rate of other periodic phenomena. represents the complex exponential function used to convert the time-domain signal into the frequency-domain, which is suitable for analyzing the vibration mode or dynamic behavior of carbon impurity particles. t represents time, which represents the real-time signal recorded by the sensor over time and reflects the time variation of particle behavior.

[0090] It should be clear that the signal processing unit does not directly monitor the vibration of carbon particles in oil. Instead, it analyzes the electrochemical response after particle adsorption through the microelectrode array sensor module. When free carbon particles adsorb onto the surface of the microelectrode array, the electrochemical signals caused by the physical properties and motion behaviors of the particles are captured by the sensor. These signals contain the characteristic information left by the movement, vibration, and collision of the particles in oil before adsorption. Through Fourier transform, the signal processing unit extracts the characteristic frequencies and intensities of particle movement and vibration from the adsorption signals to indirectly monitor the dynamic behavior of the particles.

[0091] Wavelet transform further refines the features in the signal and can distinguish carbon particles of different particle sizes. Its formula is:

[0092]

[0093] In the formula, W x (a, b) is the wavelet coefficient, representing the characteristics of carbon impurity particles in transformer oil at different scales a and time positions b. It is used to capture the local characteristics of the particles changing with time. x(t) is the sensor signal, describing the instantaneous behavior of carbon particles. The signal processing unit monitors the motion state of carbon particles in oil through optical or electrical sensors: when carbon particles move in oil, weak perturbation signals will be generated in the movement path, such as changes in the intensity of light scattering caused by flow, electric field perturbation, or tiny current fluctuations. After the sensor captures these changes, it records their position and velocity characteristics; when carbon particles gradually settle on the electrode surface, the sensor reflects the settling rate of the particles by monitoring the change in signal intensity during the deposition process; in addition, the sensor array can collect signals at different positions, and further determine the distribution state of carbon particles in the detection area by analyzing the signal distribution characteristics. The signal processing unit inputs these real-time collected sensor signals into the wavelet transform model, extracts the frequency and time characteristics of the signals through wavelet transform, and analyzes the specific behaviors of carbon particles.

[0094] ψ a,b (t) is the wavelet basis function, which is used to decompose the sensor signal at different scales a (frequency resolution) and displacements b (time positions) to highlight the local changes of carbon particles. a is the scale, representing the resolution ability for different frequency components of carbon impurity particles. Larger scales are used for low-frequency components, which are suitable for capturing the characteristics of slow movement or settling of larger particle-size carbon particles; smaller scales are used for high-frequency components, which are suitable for capturing the characteristics of rapid vibration or drastic changes of smaller particle-size carbon particles. This frequency resolution ability depends on the influence of the particle size and motion state of carbon particles on the frequency characteristics of the signal. b is the time displacement, representing the local time period in the signal, used to capture the instantaneous change characteristics of carbon particles in oil. t is the time, representing the sensor sampling moment.

[0095] In addition, a Support Vector Machine (SVM) algorithm is integrated in the signal processing unit. By quantitatively analyzing the characteristic frequency and intensity parameters, the type of carbon particles in the oil sample is judged.

[0096] The SVM algorithm trains on the collected data to establish an optimal classification model based on frequency-domain features, and extracts the resonance frequency offset Δf and the corresponding amplitude intensity I from the resonance frequency response curve of the sensor as the key classification parameters. Among them:

[0097] Δf: Represents the resonance frequency offset caused by the adsorption of carbon particles on the sensor surface. Its value is related to the equivalent particle diameter and dielectric constant of the particles (larger or more conductive particles cause a more significant Δf offset); I: Represents the attenuation intensity (dB) of the resonance signal at the characteristic frequency point, reflecting the particle concentration level (higher concentration leads to more severe signal attenuation);

[0098] Construct a two-dimensional feature space based on historical data, and determine the following category boundaries through experimental statistical distribution:

[0099] Category A (low-concentration small particles): 0 < Δf ≤ 10 kHz, I ≤ 20 dB (corresponding to non-conductive particles with a diameter d ≤ 5 μm, such as graphite debris);

[0100] Category B (high-concentration small particles): 0 < Δf ≤ 10 kHz, I > 20 dB (such as high-density submicron carbon black aggregates);

[0101] Category C (large-size conductive particles): Δf > 10 kHz, ignoring the I range (such as metallized carbon fibers, d > 20 μm).

[0102] The SVM algorithm trains on the collected data to establish an optimal classification model. The construction process of this model specifically includes the following steps: First, extract features from historical detection data to identify the key features affecting the classification results; then, according to the feature data set, by selecting an appropriate kernel function (such as the radial basis kernel function RBF or the linear kernel function) and setting appropriate parameters (such as the penalty parameter C and the parameters of the kernel function), optimize the classifier during the training process to obtain the optimal classification hyperplane or decision boundary. After normalizing Δf and I of the historical data set, determine the optimal parameter combination C = 1.5, γ = 0.8 through grid search and 5-fold cross-validation, so that the model achieves a classification accuracy of > 95% on the test set. The specific optimization of this process includes using methods such as cross-validation to adjust the model to ensure its high classification accuracy and low misclassification rate. The formula of the classification model is:

[0103]

[0104] Where f(x) is used to represent the category to which the carbon impurity particles detected by the sensor belong (for example, the type or concentration level of the particles). The result of this function reflects the classification result corresponding to the particle characteristics in the oil (such as particle size, quantity, type). sign is the sign function, and its output is positive or negative, corresponding to different categories respectively. x = [Δf, I] T represents the normalized feature vector of the input sample. K(x i , x) = exp(-γ||x i - x|| 2 ) is the RBF kernel function. When the output f(x) = +1, it can be judged that the carbon particle category belongs to category B or C (high-risk particles that need to trigger a warning). If the output f(x) = -1, it can be judged that the carbon particle category belongs to category A (normal wear particles).

[0105] The basis for classification comes from the relationship between the data characteristics collected by the sensor and the decision boundary in the classification model. The sign function gives the final classification result according to the result position of the input data point (relative to the positive and negative sides of the classification boundary). is the weighted sum, representing the weighted sum of the support vectors. α i and y i represent the weight and class label of the support vector respectively. The support vector specifically refers to the specific sample points near the classification boundary in the training stage. These sample points play a key role in the position and shape of the classification boundary. Therefore, in the model prediction stage, these support vectors affect the classification result of the new detection data through weighted summation. It reflects the similarity between the sensor data and the training samples (i.e., historical detection data) and is used to judge the new detection data. The classification boundary among them is the decision hyperplane or boundary learned by the SVM algorithm through the training data, which is used to distinguish data points of different categories. In the present invention, the classification boundary is obtained by optimizing the classification error function in the SVM model and is defined according to the characteristics of the historical data. K(x i , x) is the kernel function, calculating the input sample x and the support vector x iThe similarity between them is used to classify new carbon particle detection data through a support vector machine. As the core algorithm part of the present invention, the support vector machine is mainly used to classify carbon particle data. Specifically, the support vector machine finds the key support vectors through the training process and constructs a classification model to predict new data. ω represents the bias term, which adjusts the classification boundary to ensure the accuracy of the classifier output. The classifier is a key component in the present invention and is used to classify carbon particles based on the data collected by the sensor. Specifically, the classifier is based on the SVM model and combines the characteristics of the training data to determine the classification of carbon particles in the oil sample. In the present invention, the optimization of the bias term ω is achieved by minimizing the classification error function during the training process. By adjusting the value of the bias term ω, the specific position of the classification boundary can be adjusted, thereby optimizing the accuracy of the classification result. This optimization can effectively fine-tune the position of the decision boundary to maximize the classification accuracy in the training sample space and reduce the possibility of misclassification. The decision boundary is the boundary in the support vector machine used to distinguish different category data points. In the present invention, the decision boundary is obtained by learning the training data and is usually represented as a hyperplane in the SVM model, which divides the data space into different regions. The optimization of this boundary is achieved by adjusting the bias term ω and the weights of the support vectors in the model to better distinguish the concentration and type of carbon particles. The optimization of the bias term is achieved by minimizing the classification error function during the training process, and its role is to ensure that the decision boundary of the classifier can accurately divide the carbon particle categories in the high-dimensional feature space.

[0106] The communication module uses a DTM-100 wireless data transmission module, which supports 4G and NB-IoT communication protocols and can transmit the processed detection data to the remote monitoring platform in real time to ensure the timeliness and reliability of data transmission.

[0107] The present invention also provides a method for detecting free carbon particles in transformer oil based on a microelectrode array. The method is the usage method of the aforementioned method for detecting free carbon particles in transformer oil based on a microelectrode array, including:

[0108] The oil sample flows into the microelectrode array sensor module through the sampling module, and the sensor module adsorbs free carbon particles in the oil through the graphene oxide coating on the surface. When the carbon particles adhere to the electrode surface, it will cause changes in conductivity, and these changes are analyzed through the signal processing algorithm in the microprocessor, and finally the carbon particle concentration information is output. The processed data is transmitted to the remote monitoring platform through the communication module for real-time monitoring. When the concentration of carbon particles in the oil is detected to exceed the predetermined threshold, the alarm unit will automatically issue an alarm to prompt the staff to perform maintenance operations.

[0109] Example 1

[0110] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described in this application are only a part of the embodiments of the present invention, rather than all embodiments. Based on the spirit of the present invention, other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0111] Embodiment 1 of the present invention provides a specific implementation manner of an on-line detection device for carbon particles in transformer oil based on a microelectrode array. This device can effectively realize the on-line real-time monitoring of free carbon particles in transformer oil, and has the advantages of high precision, low cost, and easy maintenance, and is preferably applicable to the condition monitoring and fault warning of transformers.

[0112] As Figure 1 shown, Embodiment 1 of the present invention provides an on-line detection system for carbon particles in transformer oil based on a microelectrode array, which specifically includes a sampling device, a microelectrode array sensor module, a signal processing unit, a communication module, and an alarm unit.

[0113] The sampling device includes an oil sampling port arranged in the transformer oil system, and the oil sampling port is connected to the microelectrode array sensor 7 through an oil inlet pipe 4. Specifically, the oil inlet pipe 4 is preferably made of oil-resistant material to ensure no pollution when the oil sample flows into the sensor, and an electrostatic elimination device and a coarse filtration device are integrated in the oil inlet pipe 4 to prevent impurities and charges from interfering with the accuracy of detection. The flow control system of the sampling device includes a flow meter, which can realize the real-time monitoring of the oil sample flow rate and ensure the flow rate is stable.

[0114] As Figure 2 shown, the microelectrode array sensor module preferably adopts nanoscale electrode technology. The diameter of each nanoelectrode 2 is 50 nanometers, and the spacing is 50 - 100 nanometers. The electrode surface is coated with graphene oxide material. The electrodes are arranged on the sensor substrate and connected to the signal processing unit through leads. The graphene oxide layer can effectively enhance the adsorption of free carbon particles in the oil sample, thereby causing a change in conductivity. This change is transmitted to the signal processing unit through the detection signal of the microelectrode sensor.

[0115] Preferably, the nanoelectrode 2 is preferably a nanoscale metal electrode.

[0116] Preferably, the signal processing unit mainly includes Fourier transform and wavelet transform algorithm modules for analyzing the frequency characteristics of the sensor signal. Fourier transform can convert the conductivity signal from the time domain to the frequency domain and extract the characteristic frequency information, while wavelet transform can further distinguish carbon particles of different particle sizes and filter out signal noise. The data output by the signal processing unit will be transmitted to the remote monitoring platform through the communication module.

[0117] Preferably, the communication module supports 4G and NB-IoT communication protocols and is used to transmit carbon particle concentration data to the host computer in real time to ensure the timeliness and accuracy of detection information. When the detected carbon particle concentration exceeds the preset threshold, the alarm unit issues an alarm to remind the operator to maintain the transformer oil.

[0118] Specifically, the working principle of the microelectrode array sensor module is as follows: when carbon particles in the transformer oil enter the sensor module, the particles are adsorbed on the surface of the microelectrodes coated with graphene oxide, resulting in a change in conductivity. By detecting the conductivity change in real time, the sensor can quickly and accurately obtain the concentration of carbon particles in the oil sample. The signal processing unit analyzes and processes the detection signal and finally outputs the carbon particle concentration data. The entire detection process does not require the transformer to stop running, enabling online detection and greatly improving the efficiency and accuracy of carbon particle monitoring in the oil.

[0119] The detection system of the present invention has been actually tested in the transformer oil operation circulation system. The experimental results show that the system can accurately detect carbon particles in the oil within the concentration range of 1 - 10 mg / L and can also work stably at a concentration as low as 0.1 mg / L. In addition, by receiving and processing the detection data in real time through the host computer, it is possible to continuously monitor the change in the concentration of carbon particles in the transformer oil and timely discover possible fault risks.

[0120] The detection system of the present invention has a compact structure and is easy to install. It is preferably applicable to the condition monitoring of various types of transformer oils, especially having significant advantages in accurately detecting low-concentration carbon particles in the oil.

[0121] Example 2:

[0122] This Example 2 provides an improved design of an on-line detection device for carbon particles in transformer oil based on a microelectrode array. Preferably, on the basis of Example 1, the detection sensitivity, anti-interference ability and remote monitoring function of the device are further improved. The specific implementation methods are as follows.

[0123] As Figure 3 shown, the sampling device in this embodiment preferably adopts a multi-stage filtration system to ensure higher purity of the oil sample entering the sensor module, thereby improving the detection accuracy. Specifically, the sampling device includes a filter. The primary filter preferably adopts a metal mesh structure, which can effectively remove larger particle impurities in the oil sample and reduce the interference of large-particle-size impurities on the detection. Preferably, the oil inlet pipe 4 is made of a high-molecular oil-resistant material and has the characteristics of high temperature resistance and corrosion resistance, and can remain stable in the high-temperature working environment of the transformer for a long time.

[0124] Preferably, the static electricity elimination device is further optimized. It is preferably to use an electrostatic shielding layer to wrap the oil inlet pipeline 4 to prevent the detection signal from being affected by static electricity accumulation during the flow of the oil sample. The entire sampling process monitors the oil sample flow through a flowmeter to ensure that the flow rate of each sampling remains stable, so as to improve the repeatability and accuracy of the detection.

[0125] Preferably, the diameter of the sensor electrode is reduced to 30 nanometers, and the electrode spacing is reduced to the range of 40 - 80 nanometers, so as to respond more sensitively to the tiny carbon particles in the transformer oil. The graphene oxide material coated on the electrode surface has been chemically modified, which can significantly improve its adsorption capacity for carbon particles of different particle sizes. Especially when detecting carbon particles with a low concentration (less than 0.05 mg / L), it shows a more significant sensitivity improvement. The number of microelectrodes in the sensor array is also appropriately increased, preferably several hundred independent electrodes, arranged on a large-area substrate to increase the detection coverage area and detection efficiency.

[0126] Preferably, the signal processing unit integrates more complex signal processing algorithm modules, preferably including triple algorithms of adaptive filtering, Fourier transform, and wavelet transform. Specifically, the adaptive filtering algorithm is used to eliminate the noise interference in the detection signal in real time. The Fourier transform converts the conductivity signal detected by the sensor from the time domain to the frequency domain to extract the characteristic signals of carbon particles at different frequencies. The wavelet transform is further used for fine analysis of the detection signal. Preferably, it can distinguish the characteristic frequency signals of carbon particles with different particle sizes and accurately filter out the environmental noise interference through algorithms to ensure the accuracy of the detection results.

[0127] The signal processing unit preferably uses a high-performance processing chip, which can realize real-time analysis and processing of a large amount of signal data, and transmit the analysis results to the remote monitoring platform in the form of data packets through the communication module. Different from Embodiment 1, in this embodiment, the communication module preferably supports 5G and NB-IoT communication protocols. Specifically, 5G communication is used for a large amount of data transmission in high-bandwidth scenarios, while NB-IoT is suitable for stable data transmission in low-power and low-bandwidth scenarios, ensuring that the detection data can be transmitted in a timely and reliable manner in various scenarios.

[0128] Preferably, the alarm unit in this embodiment preferably introduces a dynamic alarm mechanism. Specifically, the alarm unit dynamically adjusts the alarm threshold according to the detected carbon particle concentration. This dynamic adjustment mechanism is based on a comprehensive analysis of historical detection data and real-time detection data. By setting a basic alarm threshold and combining the real-time change trend of the detection data, the adjusted alarm threshold is dynamically calculated by an algorithm. When the change rate of the carbon particle concentration is high (indicating a rapid increase in the particle concentration), the system automatically lowers the alarm threshold, thereby sending out a warning signal in advance to remind the operator to make necessary maintenance preparations; when the carbon particle concentration approaches the threshold, the system preferably sends out a warning signal to remind the operator to make necessary maintenance preparations; and when the concentration exceeds the set alarm threshold, the system will trigger a forced alarm, requiring the operator to immediately take maintenance measures to prevent serious failures caused by the deterioration of transformer oil. The core of the dynamic alarm mechanism lies in combining the absolute value and the change rate of the carbon particle concentration to dynamically adjust the threshold, thereby improving the accuracy and timeliness of the alarm.

[0129] Preferably, this embodiment is also preferably equipped with a cloud data analysis platform, which can store and analyze historical detection data for a long time, specifically including the prediction of the change trend of carbon particle concentration and the establishment of a fault warning model. Specifically, the cloud platform first performs data cleaning and feature extraction on the uploaded historical data and real-time detection data, and extracts the key features related to the carbon particle concentration in the data (such as the change rate of concentration, the fluctuation range, the detection ambient temperature, etc.). Based on the extracted features, the platform uses big data analysis techniques and machine learning algorithms (such as time series prediction models, regression analysis algorithms, or long short-term memory networks LSTM) to establish a trend model for the change of carbon particle concentration. Through the dynamic analysis of real-time data and historical data, the cloud platform can predict the change trend of carbon particle concentration in a certain period of time in the future, and generate a prediction report in real time, providing critical warning information. When the prediction report shows that the carbon particle concentration may continue to rise and approach or exceed the alarm threshold in the future, the system will actively send an early warning to the operator to help them judge the potential risks of the transformer oil state in advance and take necessary countermeasures. At the same time, the cloud platform can also establish a fault warning model based on long-term data analysis. By discovering the correlation between the change of carbon particle concentration and actual equipment failures, the dynamic adjustment mechanism of the alarm unit is further optimized to ensure more accurate and efficient monitoring and prediction.

[0130] Experimental verification shows that the detection system of this embodiment has significant detection advantages within a wider range of carbon particle concentrations (0.05 mg / L to 10 mg / L). Especially in the case of low concentrations (such as 0.05 mg / L), it can still achieve accurate monitoring of carbon particle concentration, showing higher detection sensitivity and reliability. In addition, the dynamic alarm and cloud data analysis functions of the system greatly improve its intelligence and reliability in practical applications. The following is an example of the experimental data in tabular form, covering the carbon particle concentration range (0.05 mg / L to 10 mg / L) and its detection results, verifying the sensitivity and reliability of the system:

[0131]

[0132]

[0133] In the above table, Carbon particle concentration (mg / L): The actual carbon particle concentration set in the experiment. Detection value (mg / L): The numerical value of the carbon particle concentration detected by the system. Detection error (%): The relative error between the system detection value and the actual concentration, reflecting the detection accuracy of the system. Sensitivity (mg / L): The minimum resolution ability of the system at the corresponding concentration. Reliability (%): The stability of the system detection result, which is the reliability percentage calculated based on multiple experiments. Preferably, on the basis of Embodiment 1, this embodiment further optimizes the sampling device, microelectrode array sensor module, signal processing unit, and communication and alarm module, with higher detection sensitivity and data transmission capabilities, and is particularly suitable for real-time online detection of low-concentration carbon particles in transformer oil. It also improves the remote monitoring and fault warning capabilities of the transformer operating status through the cloud analysis platform.

[0134] The present disclosure may be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for causing a processor to implement various aspects of the present disclosure.

[0135] A computer-readable storage medium can be a tangible device that can hold and store instructions for use by an instruction execution device. A computer-readable storage medium may be, for example—but not limited to—an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as a punched card or raised structures in grooves having instructions stored thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as an instantaneous signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0136] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include a copper transmission cable, an optical fiber transmission, a wireless transmission, a router, a firewall, a switch, a gateway computer, and / or an edge server. A network adapter or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0137] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state-setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer-readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field-programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. An online detection system for free carbon particles in transformer oil based on a microelectrode array, comprising: Sampling module, microelectrode array sensor module, signal processing unit, communication module, remote monitoring platform and alarm unit; characterized in that: The sampling module includes an oil inlet pipeline and a filter device. The oil sample is transported from the transformer oil tank to the microelectrode array sensor module through the oil inlet pipeline. The filter device is located inside the oil inlet pipeline. A filter membrane is provided inside the filter device to remove impurities in the oil sample. The microelectrode array sensor module includes a microelectrode array sensor, which is a nanoelectrode array arranged on a silicon substrate. The surface of the microelectrode array sensor is coated with graphene oxide to adsorb free carbon particles in the oil sample. When the free carbon particles are adsorbed to the surface of the microelectrode array sensor, an adsorption signal is generated; The signal processing unit is used to extract the characteristic frequency and intensity of the carbon particle vibration in the adsorption signal, and classify the carbon particles according to the characteristic frequency and intensity of the carbon particles.

2. The online detection system for free carbon particles in transformer oil based on microelectrode array according to claim 1, characterized in that: The sampling module also includes a servo motor, a micro pump and a valve; the valve, the micro pump and the servo motor are all located outside the electrostatic shielding layer, and the electrostatic shielding layer is coated on the outside of the oil inlet pipe; the micro pump is used to promote the flow of oil samples, the valve is used to adjust the flow rate of the oil samples, and the servo motor is used to control the opening of the valve to adjust the flow rate of the oil samples.

3. The online detection system for free carbon particles in transformer oil based on microelectrode array according to claim 2, characterized in that: The microelectrode array sensor module also includes a temperature sensor and a pressure sensor; the microelectrode array sensor, the temperature sensor and the pressure sensor are all arranged in the oil product detection chamber; the oil outlet pipeline is connected to the oil product detection chamber; the pressure sensor is connected to the servo motor through a servo motor control line.

4. The online detection system for free carbon particles in transformer oil based on microelectrode array according to claim 3, characterized in that: The microelectrode array sensor adopts a regular hexagonal honeycomb arrangement, arranging the nanoelectrodes at equal intervals on a square silicon substrate with a side length of c; from the first row of nanoelectrodes to the row with the largest number of nanoelectrodes, the distance between the leftmost and rightmost nanoelectrodes and the edge of the silicon substrate decreases row by row, with a decrease of Δd; the distance between the leftmost and rightmost nanoelectrodes in the row with the largest number of nanoelectrodes and the edge of the silicon substrate is d max ; From the row with the largest number of nanoelectrodes to the last row of nanoelectrodes, the distance between the leftmost and rightmost nanoelectrodes and the edge of the silicon substrate increases row by row, with an increase of Δd.

5. A method for online detection of free carbon particles in transformer oil based on a microelectrode array, which is implemented by using the online detection system for carbon particles in transformer oil based on a microelectrode array according to any one of claims 1 to 4, comprising: The oil sample obtained by the sampling module flows into the microelectrode array sensor module through the oil inlet pipeline; The pressure and temperature of the oil sample are monitored in real time by the pressure sensor and temperature sensor of the microelectrode array sensor module, and the micro pump and valve of the sampling module are adjusted according to the real-time change values ​​of the pressure and temperature of the oil sample to adjust the flow rate and flow velocity of the oil sample; After the flow rate and flow velocity of the oil sample are adjusted, the microelectrode array sensor generates an adsorption signal when it adsorbs free carbon particles in the oil sample; The signal processing unit extracts the characteristic frequency and intensity of particle vibration in the adsorption signal, and classifies the carbon particles according to the characteristic frequency and intensity of the carbon particles.

6. The method for online detection of free carbon particles in transformer oil based on microelectrode array according to claim 5, characterized in that: The micro pump of the sampling module is adjusted according to the real-time changes in the pressure and temperature of the oil sample. The flow rate of the micro pump after adjustment satisfies the following relationship: In the formula, Q adjust is the flow rate after micro pump adjustment, in m 3 / s; K p is the first proportionality coefficient, and its value is 1.2; P target is the target pressure, in MPa; P actual is the actual collected oil sample pressure, in MPa; μ(T) is the viscosity of the oil sample, in Pa·s, which is determined by the temperature T, and its viscosity influence coefficient α is 0.8; ΔP is the pressure change of the oil sample, in MPa, and its influence coefficient β is 0.5; C part is the carbon particle concentration in the oil sample, in mg / L; C0 is the initial carbon particle concentration, in mg / L; γ and δ are the linear and nonlinear influence coefficients of particle concentration on flow regulation, respectively; D particle is the particle size of carbon particles, in μm; K viscosity is the coefficient related to viscosity and particle size; λ1 is the adjustment coefficient of the effect of particle size on viscosity; ζ is the nonlinear influence coefficient of particle concentration on flow regulation.

7. The method for online detection of free carbon particles in transformer oil based on microelectrode array according to claim 6, characterized in that: The valve of the sampling module is adjusted according to the real-time changes in the pressure and temperature of the oil sample. The valve opening after adjustment satisfies the following relationship: In the formula, α adjust is the valve opening after adjustment, unit: °, α0 is the initial valve opening, unit: °, γ1 is the influence coefficient of temperature on valve opening, its value is 0.3, ΔT is the temperature change, unit: ℃; γ2 is the influence coefficient of pressure change on valve opening, its value is 0.4, ΔP is the pressure change of oil sample, unit: MPa; C part is the carbon particle concentration in the oil sample, unit: mg / L, C0 is the initial carbon particle concentration, unit: mg / L, β1 is the coefficient of influence of particle concentration on valve opening, its value is 0.5, η is the nonlinear response coefficient of particle concentration, its value is 1.2; Q adjust The flow rate after micro pump adjustment, unit: m 3 / s,Q max is the maximum flow rate, unit: m 3 / s, refers to the flow rate when the valve opening is adjusted to the maximum; v D is the secondary influence coefficient of flow rate on valve opening; particle is the particle size of carbon particles, unit: μm, α2 is the influence coefficient of particle size on valve opening; γ3 is the response coefficient of particle concentration to valve opening, and k is the adjustment coefficient of Sigmoid function.

8. The method for online detection of free carbon particles in transformer oil based on microelectrode array according to claim 7, characterized in that: Adjusting the flow rate and volume of the oil sample also includes: Combining PID control and machine learning algorithms for adjustment, the control output satisfies the following relationship: Among them, u(t) is the control output, the adjustment amount of the micro pump flow or valve opening; e(t) is the error signal, the difference between the target pressure and the actual pressure, t is the oil sample detection time; K c is the second proportionality coefficient, its value is 0.8; K i is the integral coefficient, its value is 0.5; K d is the differential coefficient, and its value is 0.3; ω1 is the adjustment coefficient of flow rate to PID feedback; Q actual is the actual flow rate, unit: m 3 / s,Q max is the maximum flow rate, unit: m 3 / s; ω2 is the adjustment coefficient of particle concentration to PID feedback; C part is the carbon particle concentration in the oil sample, unit: mg / L, C0 is the initial carbon particle concentration, unit: mg / L; μ is the nonlinear response coefficient of particle concentration to detection sensitivity.

9. The method for online detection of free carbon particles in transformer oil based on a microelectrode array according to claim 8, characterized in that: The signal processing unit uses Fourier transform algorithm and wavelet transform algorithm to analyze and process the adsorption signal, converts the time domain signal detected by the microelectrode array sensor into a frequency domain signal, extracts the characteristic frequency and intensity of the free carbon particles, the characteristic frequency includes the resonant frequency offset of the carbon particles, and the intensity includes the resonant signal attenuation intensity of the carbon particles, uses the wavelet transform algorithm to distinguish carbon particles of different particle sizes, and uses the SVM algorithm to classify the data to determine the content of carbon particles in the oil sample.

10. The method for online detection of free carbon particles in transformer oil based on microelectrode array according to claim 9, characterized in that: The signal processing unit classifies the carbon particles according to the characteristic frequency and intensity of the carbon particles, including: The carbon particles are classified according to the resonant frequency offset Δf of the adsorption signal and the resonant signal attenuation intensity I. When 0<Δf≤10kHz and I≤20dB, the carbon particles are judged as small particles with low concentration; when 0<Δf≤10kHz and I>20dB, the carbon particles are judged as small particles with high concentration; when Δf>10kHz, the carbon particles are judged as large-sized conductive particles.

Images