Natural ester insulating oil aging detection method and device and computer program product

Abstract

The invention discloses a natural ester insulating oil aging detection method and device and a computer program product, and the method comprises the steps: S1, synchronously obtaining aging indication data of natural ester insulating oil on line; s2, carrying out standardization processing on the aging indication data and extracting key feature parameters; s3, inputting the key characteristic parameters into an aging evaluation model trained by a deep neural network, and outputting the aging probability and risk level of the oil sample; and S4, an early warning or oil change instruction is triggered according to the risk level and a preset threshold value. According to the invention, early warning and oil change instruction triggering can be carried out in advance without power failure, so that the operation reliability and the maintenance economy are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of transformer insulation testing technology, specifically to a method, apparatus, and computer program product for testing the aging of natural ester insulating oil. Background Technology

[0002] In traditional transformer equipment, mineral insulating oil is widely used due to its excellent insulation and cooling properties and low price. In recent years, with increasing attention to environmental issues, natural ester insulating oil has gradually replaced mineral insulating oil as the insulating and cooling medium for transformers.

[0003] The main component of natural esters is triglycerides, whose molecular structure contains a large number of unsaturated bonds. During long-term operation, these bonds are prone to oxidation upon contact with air, leading to oil aging. Aging products mainly include acidic substances and sludge: acidic substances corrode the internal metal structural components of the transformer, reducing mechanical strength; sludge easily deposits on the core, winding surfaces, and radiator inner walls, hindering heat dissipation and reducing insulation performance. Furthermore, triglycerides decompose during aging, releasing fatty acid segments, further accelerating the deterioration of oil performance.

[0004] Currently, monitoring the aging status of natural esters mainly relies on offline sampling and analysis. Its limitations include: the need to shut down transformers, affecting the continuous operation of the power grid; and the lag in responding to early aging characteristics (such as microbial metabolites and accumulation of polar impurities) by traditional detection indicators (acid value, dielectric loss, breakdown voltage, etc.), making it difficult to provide early warnings of aging trends. Summary of the Invention

[0005] The technical problem to be solved by the embodiments of the present invention is to provide a method, device and computer program product for detecting the aging of natural ester insulating oil, so as to realize real-time, non-invasive diagnosis and early warning of the aging state of the oil.

[0006] To solve the above technical problems, the present invention provides a method for detecting the aging of natural ester insulating oil, comprising: Step S1: Acquire aging indicator data of natural ester insulating oil online synchronously; Step S2: Standardize the aging indicator data and extract key feature parameters; Step S3: Input the key feature parameters into the aging assessment model trained by the deep neural network, and output the aging probability and risk level of the oil sample. Step S4: Trigger an early warning or oil change command based on the risk level and preset threshold.

[0007] Preferably, in step S1, the spectral characteristics, electrochemical response, and biological metabolite data of the oil sample are simultaneously collected by a spectral analysis module, an electrochemical sensor module, and a biomarker detection module to obtain the aging indicator data.

[0008] Preferably, the spectral analysis module includes: The near-infrared spectroscopy unit is used to detect the characteristic absorption peaks of oxidation products in oil. Among them, the intensity of the hydroxyl absorption peak and the carbonyl absorption peak are positively correlated with the degree of aging. The ultraviolet-visible spectroscopy unit is used to identify absorption changes in aromatic compounds; Raman spectroscopy units are used to capture the molecular vibrational modes of fatty acids and alcohol metabolites produced by ester bond breaking.

[0009] Preferably, the electrochemical sensor module includes: A conductivity sensor used to measure the conductivity of oil samples using a four-electrode method; pH sensor is used to detect dynamic changes in the acid value of oil; Dielectric constant sensor used to assess the content of polar impurities in oil through high-frequency dielectric response.

[0010] Preferably, the biomarker detection module includes: The microbial metabolite analysis unit uses gas chromatography-mass spectrometry to quantitatively analyze volatile organic compounds produced by microbial metabolism in oil. The enzyme activity detection unit uses the fluorescent substrate method to quantify esterase activity.

[0011] Preferably, step S2 specifically includes: Perform normalization on the spectral data; Z-score normalization was performed on the electrochemical data; Perform logarithmic transformation on biological metabolite data; The intensity and peak width of the hydroxyl and carbonyl absorption peaks, the rate of change in conductivity, the increment of dielectric loss, the concentration of volatile organic compounds, and the esterase activity were extracted as the key characteristic parameters.

[0012] Preferably, the aging assessment model in step S3 is a CNN-LSTM hybrid network or a Transformer architecture, and is trained using transfer learning and adversarial generative networks. The training dataset includes natural ester insulating oil samples in four stages: new oil, lightly aged, moderately aged, and heavily aged.

[0013] Preferably, step S4 specifically includes: The risk level is compared with a preset two-level threshold. When the risk level reaches the warning threshold, a warning signal is issued; When the risk level reaches the oil change threshold, an oil change instruction is automatically generated.

[0014] The present invention also provides a natural ester insulating oil aging detection device, comprising: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the natural ester insulating oil aging detection method.

[0015] The present invention also provides a computer program product, including computer instructions that instruct a computer device to perform an operation corresponding to the method.

[0016] The present invention offers the following advantages: It employs simultaneous online sampling via three modules—spectroscopy, electrochemistry, and biomarkers—to input key features such as hydroxyl / carbonyl absorption peaks, conductivity change rates, and volatile organic compound concentrations into a CNN-LSTM / Transformer model enhanced with transfer learning and GAN in real time, enabling non-invasive, second-level assessment of the aging probability of natural ester insulating oil. The fluorescence substrate method for quantifying esterase activity significantly improves sensitivity compared to colorimetric methods, and the four-electrode method, after eliminating contact resistance, simultaneously enhances conductivity monitoring accuracy. Furthermore, the present invention can provide early warnings and trigger oil change commands without power interruption, significantly improving operational reliability and maintenance economy. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart of an aging detection method for natural ester insulating oil according to Embodiment 1 of the present invention.

[0019] Figure 2 This is a schematic diagram of the process for measuring conductivity using the four-electrode method in an embodiment of the present invention. Detailed Implementation

[0020] The following description of the embodiments is taken with reference to the accompanying drawings, which illustrate specific embodiments in which the invention can be implemented.

[0021] Please refer to Figure 1 As shown, Embodiment 1 of the present invention provides a method for detecting the aging of natural ester insulating oil, comprising: Step S1: Acquire aging indicator data of natural ester insulating oil online synchronously; Step S2: Standardize the aging indicator data and extract key feature parameters; Step S3: Input the key feature parameters into the aging assessment model trained by the deep neural network, and output the aging probability and risk level of the oil sample. Step S4: Trigger an early warning or oil change command based on the risk level and preset threshold.

[0022] Specifically, in this embodiment of the invention, step S1 involves multimodal data acquisition, using a spectral analysis module, an electrochemical sensor module, and a biomarker detection module to simultaneously acquire spectral characteristics, electrochemical responses, and bioproduct data of natural ester insulating oil.

[0023] As an example, a flow-through sample cell is installed at the transformer's online monitoring interface, allowing the oil sample to flow continuously through the cell. The spectral analysis module includes a near-infrared spectral unit, a UV-Vis spectral unit, and a Raman spectral unit, wherein: The near-infrared spectroscopy unit was used to scan oil samples in the wavelength range of 700–2500 nm and record hydroxyl groups (approximately 3400 nm). ) and carbonyl group (approximately 1710) Changes in the intensity and shape of the absorption peak; The UV-Vis spectroscopy unit is used to scan in the 200–800 nm range to identify absorption changes in furfural aromatic degradation products; Raman spectroscopy units are used in wavenumbers 100–4000 The molecular vibration spectrum is collected internally to capture the characteristic bands of free fatty acids and alcohol metabolites after ester bond breakage.

[0024] The three sub-units mentioned above are triggered synchronously to ensure that the spectral data of the same oil sample is collected completely at the same time.

[0025] The electrochemical sensor module includes a conductivity sensor, a pH sensor, and a dielectric constant sensor, which are arranged sequentially in the same oil circuit adjacent to the spectral analysis module.

[0026] The conductivity sensor adopts a four-electrode structure. A constant DC current is passed through the two outer electrodes, and the voltage difference is measured by the two inner electrodes. The conductivity is calculated based on the current, voltage difference, voltage electrode spacing, and cross-sectional area of ​​the oil sample. The pH sensor directly contacts the oil sample and outputs a dynamic acid value curve with a resolution of 0.01 pH units in real time. A dielectric constant sensor is used to apply a high-frequency electric field of 1kHz–1MHz to record the dielectric constant and dielectric loss in order to assess the content of polar impurities.

[0027] The sensor array is linked with the oil circuit temperature control component to maintain a constant oil sample temperature (25±0.5 ℃) to eliminate temperature drift.

[0028] Furthermore, such as Figure 2 As shown, the four-electrode method for measuring conductivity includes the following steps: First, configure the electrodes: use four inert metal electrodes (such as platinum or gold electrodes), with two pairs of electrodes used to apply current (current electrodes) and measure voltage (voltage electrodes) respectively. At the same time, the distance between the current electrodes should be greater than the distance between the voltage electrodes (for example, the distance between the current electrodes is 20 mm and the distance between the voltage electrodes is 10 mm) to reduce the influence of contact resistance. Then apply current: apply a constant DC current (range 0.1-10 mA) to the oil sample through the current electrode to avoid overcurrent that could cause sample decomposition; Next, measure the voltage: use a high-precision voltmeter to measure the voltage difference between the voltage electrodes. ; Finally, the conductivity was calculated: Based on the current, voltage difference, voltage electrode spacing, and cross-sectional area of ​​the oil sample, the conductivity was calculated. The calculation is as follows:

[0029] in, For current, The distance between voltage electrodes. This represents the cross-sectional area of ​​the oil sample.

[0030] It should be noted that the conductivity changes significantly with temperature and must be operated in a constant temperature chamber (e.g., 25±0.5℃). Therefore, calibration and correction steps are also included, namely, verifying the accuracy of the equipment and correcting the contact resistance by measuring the conductivity of a standard solution (e.g., KCl solution).

[0031] Oil samples are continuously drawn from the bypass micro-sampling port of the same oil circuit and processed in two paths through the biomarker detection module: Microbial metabolite analysis pathway: Quantitative analysis using gas chromatography-mass spectrometry to detect volatile organic compounds produced by microbial metabolism (such as fungi and bacteria) in oil; Enzyme activity detection pathway: After diluting the oil sample, a fluorescent substrate is added, and the mixture is shaken under set conditions for a period of time. The esterase activity is quantified by the change in fluorescence intensity.

[0032] The three data streams (spectral, electrochemical, and biological markers) are timestamped by the system clock, achieving true synchronous acquisition and providing a time-series consistency guarantee for subsequent data fusion.

[0033] In a specific implementation process, quantitative analysis using gas chromatography-mass spectrometry includes the following steps: (1) Oil sample preparation: Weigh 0.1g of insulating oil sample, add 3g NaCl (to increase ionic strength) and 10mL of ultrapure water, mix well and place in headspace vial, add 20µL of decane-D25 to correct quantitative error; (2) Extraction and desorption: headspace extraction at 60℃ for 40 min (adjust the time according to the volatility of the target substance), SPME fiber is inserted into the gas chromatograph injection port (250℃) for thermal desorption for 5 min, so that the adsorbate is desorbed and enters the gas chromatograph column; (3) Gas chromatography-mass spectrometry: High-purity helium (99.999%) was introduced into the gas chromatograph at a flow rate of 1 mL / min and the initial temperature was 40℃ for 10 min. Then, the temperature was increased to 260℃ at 10℃ / min and held for 10 min. The mass spectrometer was ionized by electron bombardment with an ion source temperature of 230℃ and a scanning range of m / z 35-500. (4) Calculate the concentration of the target substance using the internal standard method.

[0034] In a specific implementation process, the fluorescent substrate method for detecting enzyme activity is suitable for quantifying esterase activity in natural ester insulating oils (such as the rate of ester bond decomposition by microorganisms in the oil). The specific operating steps are as follows: (1) Substrate selection and preparation: Select esterase-specific substrates or AMC (7-amino-4-methylcoumarin) labeled substrates; (2) Solution preparation: Solution 1: AMC substrate dissolved in DMSO (dimethyl sulfoxide) to a final concentration of 500 µmol / L; Solution 2: pH 7.4 phosphate buffer; (3) Diluting the sample: Dilute the insulating oil sample 1:50 and homogenize it in an ice bath for 3 min; (4) Dispensing reaction: Add 200 µL of diluent to the microplate and add 50 µL of substrate solution to each well; (5) Reaction start-up: Incubate at 30℃ with constant temperature shaking for 3 hours (700 rpm); (6) Fluorescence detection: excitation wavelength 355 nm, emission wavelength 460 nm (AMC substrate), record the initial and final fluorescence intensities, and calculate the ΔF value; (7) Data analysis: Use AMC standards of known concentration to plot fluorescence intensity-concentration curves.

[0035] Step S2 involves fusing and extracting features from the collected data, including standardization, and extracting the intensity of spectral absorption peaks, the rate of change of conductivity, and the concentration of volatile organic compounds as key feature parameters.

[0036] Data standardization includes the following: Spectral data were normalized (0-1 range) to eliminate device differences; Electrochemical data were standardized using Z-score (mean 0, standard deviation 1). Biomarker data are reduced for skewed distributions through logarithmic transformation (log10).

[0037] It should be noted that the characteristics of spectral absorption peak intensity, conductivity change rate, and volatile organic compound concentration extracted in step S2 include the following: Spectral characteristics: Extracted hydroxyl absorption peak (3400) ), carbonyl absorption peak (1710) The intensity and peak width of ( ); Electrochemical characteristics: Calculate the rate of change of conductivity and the increment of dielectric loss; Biological characteristics: quantification of volatile organic compound concentrations (such as ethanol and methane) and esterase activity (μmol / min / mg).

[0038] Step S3 utilizes a pre-built aging assessment model trained with a deep neural network to output the aging probability and risk level of the oil sample. Specific details are as follows: The key characteristic parameters obtained in step S2 (hydroxyl absorption peak intensity, carbonyl absorption peak intensity, conductivity change rate, dielectric loss increment, volatile organic compound concentration, and esterase activity) are fed in real time into the aging assessment model that has been trained offline. The aging assessment model is as follows: Network architecture: A CNN-LSTM hybrid network (CNN layers perform local convolutions on spectral images and electrochemical curves to extract spatial features; LSTM layers receive the output of CNN layers and unfold them step-by-step to capture the long-term dependencies of the same oil sample within a continuous monitoring period), or a Transformer architecture (using a multi-head self-attention mechanism to directly model multimodal feature sequences and capture the global correlation between spectral, electrochemical, and biological features). Both have been pre-trained on a mineral oil dataset using transfer learning and fine-tuned using natural ester samples with four aging levels: "new oil," "lightly aged," "moderately aged," and "severely aged." Among them, the new oil samples are data from unaged natural ester insulating oil; the lightly aged samples are data from oil samples after accelerated aging tests; and the moderately or severely aged samples are data from oil samples replaced during actual operation. In addition, an adversarial generative network (GAN) is used to synthesize extreme aging data based on the severely aged samples to expand the training set distribution boundary and improve the robustness of the model under abnormal operating conditions. Input dimension: 6-dimensional feature vector (the above 6 key feature parameters); Output: The aging probability (0–1 continuous value) and risk level (discrete 0–3 level) of the current oil sample.

[0039] It has been verified that the hydroxyl absorption peak (approximately 3400 nm) in the near-infrared spectrum... ) and carbonyl absorption peak (approximately 1710) The strength of the aging level increases monotonically with increasing aging level. This positive correlation is incorporated as prior knowledge into the constraint term of the loss function to improve the model's convergence speed.

[0040] Model inference is completed in a single forward propagation within the edge computing unit, and the output is directly used for subsequent early warning or oil change decisions.

[0041] Step S4 triggers an early warning or oil change command based on the risk level and a preset threshold. One feasible implementation is to send the risk level (level 0–3) output by the model to the threshold determination unit in real time. The system has two preset threshold levels: Warning threshold = Level 2, triggering a "yellow" alarm signal and uploading it to the monitoring master station; When the oil change threshold is level 3, a "red" alarm is triggered and an oil change work order is automatically generated.

[0042] Alarm information is delivered to the maintenance terminal within a set time through the station's communication protocol, enabling uninterrupted power supply reminders and maintenance decisions.

[0043] Corresponding to the natural ester insulating oil aging detection method described in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides a natural ester insulating oil aging detection device, comprising: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the natural ester insulating oil aging detection method described in Embodiment 1 of the present invention.

[0044] Corresponding to the natural ester insulating oil aging detection method described in Embodiment 1 of the present invention, Embodiment 3 of the present invention also provides a computer program product, including computer instructions, which instruct a computer device to perform the operation corresponding to the natural ester insulating oil aging detection method described in Embodiment 1 of the present invention.

[0045] Preferably, the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can be any conventional processor. The processor is the control center of the device, connecting various parts of the device through various interfaces and lines.

[0046] The memory mainly includes a program storage area and a data storage area. The program storage area can store the operating system, applications required for at least one function, etc., while the data storage area can store related data, etc. Furthermore, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard drive, a SmartMedia Card (SMC), a Secure Digital (SD) card, and a Flash Card, or other volatile solid-state storage devices.

[0047] It should be noted that the above-mentioned devices may include, but are not limited to, processors and memory, as will be understood by those skilled in the art.

[0048] As can be seen from the above description, compared with the prior art, the beneficial effects of the present invention are as follows: The present invention uses three modules—spectroscopy, electrochemistry, and biomarkers—to simultaneously sample online, inputting key features such as hydroxyl / carbonyl absorption peaks, conductivity change rate, and volatile organic compound concentration into a CNN-LSTM / Transformer model enhanced by transfer learning and GAN in real time, thereby achieving a non-invasive, second-level assessment of the aging probability of natural ester insulating oil; the fluorescence substrate method for quantifying esterase activity has significantly improved sensitivity compared to the colorimetric method, and the conductivity monitoring accuracy is simultaneously improved after the four-electrode method eliminates contact resistance; the present invention can provide early warning and trigger oil change commands under uninterrupted power conditions, significantly improving operational reliability and maintenance economy.

[0049] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for detecting the aging of natural ester insulating oil, characterized in that, include: Step S1: Acquire aging indicator data of natural ester insulating oil online synchronously; Step S2: Standardize the aging indicator data and extract key feature parameters; Step S3: Input the key feature parameters into the aging assessment model trained by the deep neural network, and output the aging probability and risk level of the oil sample. Step S4: Trigger an early warning or oil change command based on the risk level and preset threshold.

2. The method according to claim 1, characterized in that, In step S1, the spectral characteristics, electrochemical response, and biological metabolite data of the oil sample are simultaneously collected by the spectral analysis module, the electrochemical sensor module, and the biomarker detection module to obtain the aging indicator data.

3. The method according to claim 2, characterized in that, The spectral analysis module includes: The near-infrared spectroscopy unit is used to detect the characteristic absorption peaks of oxidation products in oil. Among them, the intensity of the hydroxyl absorption peak and the carbonyl absorption peak are positively correlated with the degree of aging. The ultraviolet-visible spectroscopy unit is used to identify absorption changes in aromatic compounds; Raman spectroscopy units are used to capture the molecular vibrational modes of fatty acids and alcohol metabolites produced by ester bond breaking.

4. The method according to claim 2, characterized in that, The electrochemical sensor module includes: A conductivity sensor used to measure the conductivity of oil samples using a four-electrode method; pH sensor is used to detect dynamic changes in the acid value of oil; Dielectric constant sensor used to assess the content of polar impurities in oil through high-frequency dielectric response.

5. The method according to claim 2, characterized in that, The biomarker detection module includes: The microbial metabolite analysis unit uses gas chromatography-mass spectrometry to quantitatively analyze volatile organic compounds produced by microbial metabolism in oil. The enzyme activity detection unit uses the fluorescent substrate method to quantify esterase activity.

6. The method according to claim 1, characterized in that, Step S2 specifically includes: Perform normalization on the spectral data; Z-score normalization was performed on the electrochemical data; Perform logarithmic transformation on biological metabolite data; The intensity and peak width of the hydroxyl and carbonyl absorption peaks, the rate of change in conductivity, the increment of dielectric loss, the concentration of volatile organic compounds, and the esterase activity were extracted as the key characteristic parameters.

7. The method according to claim 1, characterized in that, The aging assessment model in step S3 is a CNN-LSTM hybrid network or a Transformer architecture, and is trained using transfer learning and adversarial generative networks. The training dataset contains natural ester insulating oil samples in four stages: new oil, lightly aged, moderately aged, and heavily aged.

8. The method according to claim 1, characterized in that, Step S4 specifically includes: The risk level is compared with a preset two-level threshold. When the risk level reaches the warning threshold, a warning signal is issued; When the risk level reaches the oil change threshold, an oil change instruction is automatically generated.

9. A device for detecting the aging of natural ester insulating oil, characterized in that, include: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the natural ester insulating oil aging detection method as described in any one of claims 1 to 8.

10. A computer program product, characterized in that, Includes computer instructions that instruct a computer device to perform an operation corresponding to the method as described in any one of claims 1 to 8.

Images