A steam turbine main oil tank lubricating oil water content on-line monitoring and inversion method

Abstract

This invention provides an online monitoring and inversion method for the water content of lubricating oil in the main oil tank of a steam turbine, belonging to the field of steam turbine operation and maintenance monitoring technology. It can significantly alleviate or solve the problems of existing lubricating oil water content detection methods, such as reliance on manual sampling, detection lag, and difficulty in continuous online monitoring. The invention includes: collecting raw spectral data of the lubricating oil in the main oil tank; preprocessing the raw spectral data to obtain the spectral data to be analyzed; extracting absorbance characteristic parameters within a preset water content sensitive band; inputting the absorbance characteristic parameters into a pre-established water content inversion model to obtain the lubricating oil water content; outputting the water content, and outputting an alarm message when the water content exceeds a preset threshold. This invention enables online monitoring of the water content of lubricating oil in the main oil tank of a steam turbine, improving the timeliness and continuity of monitoring.

Description

Technical Field

[0001] This invention belongs to the field of steam turbine operation and maintenance monitoring technology, specifically relating to an online monitoring and inversion method for the water content of lubricating oil in the main oil tank of a steam turbine. Background Technology

[0002] As a key power unit in thermal power plants, the steam turbine's operational stability, peak-shaving response capability, and long-term service reliability directly affect the safety and economy of the power generation system. During turbine operation, the main oil tank lubricating oil not only lubricates bearings and cools components, but also significantly impacts friction control, wear suppression, and stable system operation between rotating parts. Therefore, the quality of the main oil tank lubricating oil has always been a key focus in turbine operation and maintenance management. Deterioration in lubricating oil quality can easily lead to deteriorated lubrication performance, abnormal wear of metal components, corrosion, jamming, and unplanned unit shutdowns, ultimately affecting the continuous and stable operation of the entire thermal power plant. Existing technical data also clearly indicates that the lubricating oil in the turbine's main oil tank deteriorates over long-term use due to oxidation, moisture contamination, impurity accumulation, and solid particulate contamination, potentially leading to equipment wear, corrosion, and shutdown risks.

[0003] Among the various quality indicators of the main oil tank lubricating oil, water content is a critical parameter. Once water mixes into the lubricating oil, it not only reduces lubrication efficiency but may also induce oil emulsification, accelerate oxidation and deterioration, and adversely affect turbine bearings and related metal components. Especially under long-term continuous operation, if the changing trend of water content in the lubricating oil cannot be monitored in a timely manner, it becomes difficult to perform timely oil treatment, oil changes, or fault warnings, easily leading to the gradual accumulation and expansion of problems. Therefore, rapid, continuous, and online monitoring and quantitative analysis of the water content of the turbine main oil tank lubricating oil has become an important technical requirement in the field of turbine operation and maintenance monitoring. Related solutions have considered using miniature spectrometers, ultraviolet-visible fiber optic transmission, and data processing and correlation model analysis to collect and process the spectral signals of the lubricating oil inside the main oil tank to achieve online detection of oil quality parameters. This indicates that using spectral information for inverse analysis of lubricating oil condition has a practical application basis.

[0004] Currently, the testing methods for lubricating oil in the main oil tank of steam turbines can be broadly categorized into two types. The first type involves sampling and sending samples for testing, where maintenance personnel periodically extract lubricating oil samples from the main oil tank and send them to a professional testing institution for analysis. The second type involves on-site testing using large-scale oil quality analyzers. While the former can obtain some test results, it suffers from long testing cycles, untimely data feedback, and an inability to reflect real-time changes in oil quality. Furthermore, secondary contamination may be introduced during sampling, transportation, and testing, affecting the representativeness and timeliness of the test results. Although the latter shortens the testing chain to some extent, the equipment is typically large and heavy, difficult to install and deploy, and has high requirements for the on-site environment and operating conditions. It often requires professional personnel for maintenance and operation, and still cannot meet the needs of long-term, continuous, and online monitoring of the main oil tank of steam turbines. Furthermore, from the perspective of "online water content inversion," existing technologies still have at least the following shortcomings: First, existing sampling and testing methods are essentially intermittent, unable to continuously track the water content of lubricating oil, thus making it difficult to reflect short-term fluctuations and abnormal changes in water content in a timely manner; Second, sampling methods usually rely on manual operation, which is not only labor-intensive but also easily affected by the external environment during the sampling process, resulting in problems such as sample contamination, changes in sample state, and delayed test results; Third, existing large-scale on-site testing equipment is not adaptable enough to the turbine main oil tank scenario, making it difficult to achieve compact structure, convenient installation, and long-term stable deployment, which is not conducive to continuous monitoring without affecting the normal operation of the unit; Fourth, existing technologies usually tend to perform general testing of oil quality, lacking a stable online inversion path for the key indicator of water content, especially in the main oil tank environment. There is still room for improvement in how to obtain effective spectral data without frequent sampling, preprocess the original spectrum, and achieve online quantitative inversion of water content based on spectral characteristics.

[0005] To address this, a method for online monitoring and inversion of the water content in the main oil tank of a steam turbine is proposed. Summary of the Invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method for online monitoring and inversion of the water content of lubricating oil in the main oil tank of a steam turbine.

[0007] This invention provides a method for online monitoring and inversion of the water content of lubricating oil in the main oil tank of a steam turbine, comprising the following steps: S1: Collect raw spectral data of the lubricating oil in the main oil tank of the steam turbine; S2: Preprocess the raw spectral data to obtain the spectral data to be analyzed; S3: Extract the absorbance characteristic parameters of the spectral data to be analyzed within a preset water-sensitive band; S4: Input the absorbance characteristic parameters into the pre-established water content inversion model to obtain the water content of the lubricating oil; S5: Output the water content of the lubricating oil, and output an alarm message when the water content of the lubricating oil exceeds a preset threshold.

[0008] Furthermore, in step S1, the raw spectral data of the lubricating oil is acquired by a spectrometer, an incident optical fiber, and a receiving optical fiber located outside the main oil tank of the steam turbine. The detection wavelength range of the spectrometer is 200nm to 1000nm, and the spectral resolution of the spectrometer is no greater than 2nm.

[0009] Specifically, an anti-pollution probe is provided at the end of the incident optical fiber, and a signal acquisition probe is provided at the end of the receiving optical fiber.

[0010] Specifically, both the pollution prevention probe and the signal acquisition probe are installed in the main oil tank of the steam turbine at a position below the minimum normal operating liquid level.

[0011] Preferably, in step S2, the preprocessing includes noise reduction processing, baseline correction processing, and normalization processing.

[0012] Specifically, in step S3, the absorbance characteristic parameter is the absorbance characteristic quantity of the spectral data to be analyzed in a preset water-sensitive band. The absorbance characteristic parameter includes one or more of the following in the preset water-sensitive band: absorbance peak value, absorbance peak area, ratio of absorbance corresponding to different characteristic wavelengths, first derivative characteristic, and second derivative characteristic.

[0013] Furthermore, in step S4, the moisture content inversion model is a model establishing the correspondence between absorbance characteristic parameters and moisture content based on lubricating oil samples with known moisture content standard values.

[0014] Furthermore, the moisture content inversion model is established, including the following steps: Standard spectral data were collected from multiple sets of lubricating oil samples with different water contents; The standard spectral data is preprocessed and the corresponding absorbance characteristic parameters are extracted; and The absorbance characteristic parameters are correlated with the known water content standard values ​​of each lubricating oil sample to establish the water content inversion model.

[0015] Furthermore, in step S5, the alarm information output is audible and visual alarm information and / or remote push alarm information.

[0016] Specifically, steps S1 to S5 are repeated at preset time intervals to monitor the water content of the lubricating oil in the turbine main oil tank online.

[0017] The beneficial effects of this invention are as follows: This invention acquires the raw spectral data of lubricating oil in the main oil tank of a steam turbine, and sequentially performs preprocessing, absorbance characteristic parameter extraction, and water content inversion model calculation. This allows for online acquisition of lubricating oil water content during equipment operation, eliminating the need for frequent manual sampling and offline testing, thus improving the timeliness and continuity of lubricating oil water content monitoring. Furthermore, this invention utilizes a miniature spectrometer located outside the main oil tank, along with incident and receiving optical fibers that work in conjunction with probes inside the main oil tank for spectral acquisition. This design offers advantages such as compact structure, convenient installation, and suitability for online deployment. In addition, this invention can output alarm information when the lubricating oil water content exceeds a preset threshold, enabling operators to promptly grasp the lubricating oil status in the main oil tank and take appropriate measures. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the steps of an online monitoring and inversion method for the water content of lubricating oil in the main oil tank of a steam turbine, according to a specific embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown in the figure, a method for online monitoring and inversion of the water content of lubricating oil in the main oil tank of a steam turbine, provided by a specific embodiment of the present invention, includes the following steps: S1: Collect raw spectral data of lubricating oil in the main oil tank of the steam turbine. The raw spectral data includes dark spectrum, reference spectrum and sample spectrum. S2: Preprocess the raw spectral data to obtain the spectral data to be analyzed; S3: Extract the absorbance characteristic parameters of the spectral data to be analyzed within the preset water-sensitive band. The absorbance characteristic parameters are obtained by performing dark spectrum subtraction on the reference spectrum and the sample spectrum and calculating the transmittance. S4: Input the absorbance characteristic parameters into the pre-established water content inversion model to obtain the water content of the lubricating oil; S5: Outputs the water content of the lubricating oil and outputs an alarm message when the water content of the lubricating oil exceeds a preset threshold.

[0021] In one embodiment, the water content of the lubricating oil in the main oil tank is acquired online by means of a spectral acquisition component, an optical fiber transmission component, and a data processing component installed outside the main oil tank of the steam turbine. By acquiring, analyzing, and retrieving the spectral information of the lubricating oil, the water content result of the lubricating oil can be obtained without frequent manual sampling, thus facilitating operators to promptly grasp the changes in the moisture content of the lubricating oil in the main oil tank.

[0022] Preferably, this method is applicable to online monitoring scenarios of lubricating oil in the main oil tank of a thermal power unit turbine, especially suitable for equipment maintenance that requires long-term continuous operation and is highly sensitive to changes in the water content of the lubricating oil. The online inversion method uses the spectral response of the lubricating oil within a preset water-sensitive band as basic data. It preprocesses the original spectral data, extracts absorbance characteristic parameters reflecting changes in water content in the lubricating oil, and combines this with a pre-established water content inversion model to quantitatively calculate the water content of the lubricating oil. To obtain the absorbance characteristic parameters used for water content inversion, the original spectral data includes a dark spectrum, a reference spectrum, and a sample spectrum. The dark spectrum is the background signal acquired by a miniature spectrometer when the detection light source is turned off, used to characterize the detector's background noise. The reference spectrum is acquired using lubricating oil under baseline conditions. The reference spectral signal; the sample spectrum is the real-time spectral signal collected from the lubricating oil to be tested in the main oil tank during online monitoring; the reference lubricating oil is preferably a standard lubricating oil sample that belongs to the same oil type, brand or base oil system as the online monitoring object and has a water content lower than the preset reference value; in one embodiment, the reference lubricating oil spectrum collected during the initial normal operation of the equipment can also be used as the reference spectrum; after obtaining the dark spectrum, reference spectrum and sample spectrum, the dark spectrum is first subtracted from the reference spectrum and the sample spectrum to obtain the corrected reference signal and the corrected sample signal; then the transmittance is calculated based on the corrected reference signal and the corrected sample signal; finally, the absorbance is calculated based on the transmittance to obtain the absorbance spectrum; the absorbance A(λ) is calculated according to the following formula: A(λ) = -log10(T(λ)); Where T(λ) is the transmittance at wavelength λ, and: T(λ)=[Is(λ)-Id(λ)] / [I0(λ)-Id(λ)]; Where Is(λ) is the sample spectral signal intensity at wavelength λ, I0(λ) is the reference spectral signal intensity at wavelength λ, and Id(λ) is the dark spectral signal intensity at wavelength λ.

[0023] Furthermore, the method can be repeated at preset time intervals to form continuous monitoring of the water content of the lubricating oil in the main oil tank of the steam turbine. When the monitoring result exceeds the preset threshold, an alarm message is issued in a timely manner so that the operators can take corresponding measures.

[0024] Based on the above basic implementation method, in step S1, the original spectral data of the lubricating oil is acquired by a spectrometer, an incident optical fiber, and a receiving optical fiber set outside the main oil tank of the steam turbine. The detection wavelength range of the spectrometer is 200nm to 1000nm, and the spectral resolution of the spectrometer is no greater than 2nm. The spectrometer is a miniature spectrometer. The miniature spectrometer is preferably installed on the outer side wall of the main oil tank of the steam turbine and connected to the outer shell of the main oil tank through a bracket to facilitate installation, disassembly, and maintenance. In a preferred embodiment, a shock-absorbing structure can also be set on the bracket to reduce the impact of the operating vibration of the main oil tank on the spectral acquisition process.

[0025] Furthermore, the miniature spectrometer includes a light source output end and a signal input end. The incident optical fiber is connected to the light source output end and is used to transmit the detection light to the lubricating oil inside the main oil tank. The receiving optical fiber is connected to the signal input end and is used to transmit the spectral response signal formed after being acted upon by the lubricating oil back to the miniature spectrometer to obtain the original spectral data of the lubricating oil.

[0026] In one specific embodiment, an anti-pollution probe is provided at the end of the incident optical fiber, and a signal acquisition probe is provided at the end of the receiving optical fiber; both the anti-pollution probe and the signal acquisition probe are located in the main oil tank of the steam turbine at a position below the minimum normal operating liquid level, so that the anti-pollution probe and the signal acquisition probe are immersed in lubricating oil during online monitoring.

[0027] In this embodiment, the anti-contamination probe is used to reduce contamination caused by impurities, sludge, or deposits in the lubricating oil directly acting on the end of the incident optical fiber. The signal acquisition probe is used to receive the spectral response signal after being acted upon by the lubricating oil. Both probes cooperate with the incident and receiving optical fibers respectively to form an optical path structure suitable for online acquisition inside the oil tank. Preferably, the probe extends into the main oil tank through a pre-reserved mounting hole. For example, in one specific embodiment, the probe can be positioned at a predetermined height from the bottom of the main oil tank to balance immersion requirements and the influence of sediment at the bottom of the tank.

[0028] Furthermore, the incident and receiving optical fibers can be ultraviolet-visible light transmission fibers, preferably with low transmission loss to improve the quality of the acquired spectral signal. In one example, the diameter of both the incident and receiving optical fibers can be 0.5 mm, and the transmission loss can be 0.3 dB / m. The anti-pollution probe and the signal acquisition probe are arranged opposite each other inside the main oil tank of the steam turbine, forming a preset detection optical path between their light-transmitting end faces. The detection light output from the incident optical fiber is injected into the lubricating oil through the anti-pollution probe, passes through the lubricating oil medium within the preset detection optical path, is received by the signal acquisition probe, and is transmitted to the miniature spectrometer via the receiving optical fiber. Thus, the anti-pollution probe and the signal acquisition probe together constitute an opposing transmission optical path for acquiring the transmission spectral signal of the lubricating oil within the preset detection optical path. The preset detection optical path can be set according to the type of lubricating oil, the detection sensitivity requirements, and the on-site installation space; preferably, under the same lubricating oil type and the same installation conditions, the preset detection optical path remains constant to improve the comparability of spectral data obtained at different sampling times; The anti-contamination probe is positioned at the end of the incident optical fiber to guide the detection light into the lubricating oil and reduce the impact of sludge, impurities, or deposits on the optical performance of the incident end. The signal acquisition probe is positioned at the end of the receiving optical fiber to receive the transmitted light signal after absorption by the lubricating oil. Both the anti-contamination probe and the signal acquisition probe are positioned below the minimum operating liquid level to ensure that the preset detection optical path remains within the lubricating oil coverage area during online monitoring. The anti-contamination probe and the signal acquisition probe are fixed to the pre-reserved mounting holes in the main oil tank via mounting bases, arranged opposite each other with their central axes roughly corresponding to form a stable transmission detection channel.

[0029] In another specific embodiment, in step S2, the preprocessing includes noise reduction processing, baseline correction processing, and normalization processing.

[0030] In this embodiment, noise reduction processing is used to reduce high-frequency random noise and interference signals in the original spectral data and during the acquisition process. Baseline correction processing is used to correct baseline drift caused by light source fluctuations, optical path changes, probe surface adhesion or other environmental factors. Normalization processing is used to eliminate the influence of overall light intensity differences between different sampling times on subsequent feature extraction and water content inversion results.

[0031] Specifically, in one embodiment, noise reduction processing can be performed by smoothing filtering on the original spectral data to improve the smoothness of the spectral curve; baseline correction processing can correct the background drift of the original spectral data to make the spectral data obtained at different sampling times more comparable; normalization processing can normalize the amplitude of the spectral data to make the spectral data to be analyzed under different operating conditions within a uniform scale range.

[0032] Furthermore, the spectral data obtained after preprocessing can better retain the spectral characteristic information corresponding to the changes in moisture in the lubricating oil, thus providing a data basis for the subsequent extraction of absorbance characteristic parameters and the use of the water content inversion model.

[0033] In another specific embodiment, in step S3, the absorbance characteristic parameter is the absorbance characteristic quantity of the spectral data to be analyzed in a preset water-sensitive band. The absorbance characteristic parameter includes one or more of the following: absorbance peak value, absorbance peak area, ratio of absorbance corresponding to different characteristic wavelengths, first derivative characteristic, and second derivative characteristic in the preset water-sensitive band.

[0034] Furthermore, the preset water-sensitive band can be determined by performing spectral tests and comparative analysis on multiple groups of lubricating oil samples with known water content. That is, the band that responds more significantly to changes in water content in the lubricating oil and has higher distinguishability is selected from the detection wavelength range. In practical applications, only one water-sensitive band can be selected, or multiple water-sensitive bands can be selected together for feature extraction. The preset water-sensitive band is determined through preliminary experiments, specifically: within the detection wavelength range of the miniature spectrometer, spectral acquisition is performed on multiple groups of lubricating oil samples with different water contents, and the results are analyzed for each group of lubricating oil samples. By comparing and analyzing the spectral curves, bands that show a significant response to changes in water content and have good discriminative power are selected as preset water content sensitive bands. The preset water content sensitive bands can be one band or multiple bands. When multiple bands are used, absorbance characteristic parameters in each band can be extracted separately and combined to form a feature vector for water content inversion. One or more of the following are extracted from the preset water content sensitive bands: absorbance peak value, absorbance peak area, ratio of absorbance corresponding to different characteristic wavelengths, first derivative feature, and second derivative feature, as input parameters of the water content inversion model.

[0035] Specifically, in one embodiment, a single absorbance feature parameter can be extracted from the spectral data to be analyzed as the input of the water content inversion model, or multiple absorbance feature parameters can be combined to form a feature vector and then input into the water content inversion model to improve the characterization ability of changes in the water content of lubricating oil; for example, the absorbance peak value, peak area and absorbance ratio can be selected simultaneously as model input parameters.

[0036] In another specific embodiment, in step S4, the moisture content inversion model is a model establishing the correspondence between absorbance characteristic parameters and moisture content based on lubricating oil samples with known moisture content standard values. When establishing the moisture content inversion model, multiple groups of lubricating oil samples with different moisture contents are selected as modeling samples. Standard spectral data are collected for each group of lubricating oil samples, and the corresponding moisture content standard values ​​are obtained. The moisture content standard values ​​are obtained through conventional offline moisture detection methods and used as reference values ​​for establishing the moisture content inversion model. After preprocessing the standard spectral data of each group of lubricating oil samples, the corresponding absorbance characteristic parameters are extracted, and the absorbance characteristic parameters are correlated with the corresponding moisture content standard values ​​to establish the moisture content inversion model. After the moisture content inversion model is established, it can be validated using lubricating oil samples that were not involved in the modeling. The applicability of the moisture content inversion model is evaluated by comparing the differences between the model output values ​​and the offline detection standard values.

[0037] Furthermore, the establishment of the moisture content inversion model can be based on multiple lubricating oil samples with different moisture content levels. By acquiring the absorbance characteristic parameters of each lubricating oil sample and correlating them with the corresponding known moisture content standard values, a quantitative relationship between the absorbance characteristic parameters and the moisture content of the lubricating oil is established. The known moisture content standard values ​​can be obtained through conventional offline detection methods; the moisture content standard values ​​for the modeling and validation samples are obtained through offline moisture detection methods. Preferably, the offline moisture detection method is the Karl Fischer moisture determination method; other offline detection methods capable of accurately determining the moisture content in lubricating oil can also be used to obtain the moisture content standard values. By using the moisture content standard values ​​obtained through offline detection as modeling reference values, a correspondence between the absorbance characteristic parameters and the actual moisture content of the lubricating oil can be established, thereby improving the quantitative inversion capability of the moisture content inversion model.

[0038] Specifically, the water content inversion model can be a fitting model or a regression model, the purpose of which is to realize the conversion from absorbance characteristic parameters to lubricating oil water content. In practical applications, the absorbance characteristic parameters obtained online and after preprocessing are input into the water content inversion model, and the corresponding lubricating oil water content result can be output.

[0039] In another specific embodiment, a moisture content inversion model is established, including the following steps: Standard spectral data of multiple lubricating oil samples with different water contents were collected; the standard spectral data were preprocessed and the corresponding absorbance characteristic parameters were extracted; and the absorbance characteristic parameters were correlated with the known water content standard values ​​of each lubricating oil sample to establish a water content inversion model; the known water content standard values ​​were obtained by conventional moisture detection methods.

[0040] Furthermore, during the model establishment process, multiple groups of lubricating oil samples can be prepared or selected according to different water content levels, and standard spectral data of each group of lubricating oil samples can be collected using the same or equivalent spectral acquisition conditions. Subsequently, the standard spectral data can be subjected to a preprocessing procedure corresponding to the online monitoring stage to ensure that the data processing path is consistent between the model establishment stage and the online application stage.

[0041] Furthermore, after extracting the absorbance characteristic parameters corresponding to each lubricating oil sample, the absorbance characteristic parameters can be fitted or correlated with the known moisture content standard values ​​of each sample to obtain a moisture content inversion model for online inversion. If necessary, additional samples can be used to verify the established model in order to determine the applicability and calculation accuracy of the model in different moisture content ranges.

[0042] In one specific implementation, in step S5, the output alarm information is audible and visual alarm information and / or remote push alarm information.

[0043] Specifically, alarm information can be output through the on-site display terminal, or displayed and pushed through a host computer, industrial tablet PC, or remote monitoring platform connected to the data processing module; the spectral analysis system can use an industrial-grade tablet PC with built-in dedicated analysis software, connected to a miniature spectrometer, to display parameter values ​​and change curves.

[0044] Furthermore, a lubricating oil moisture content alarm threshold can be preset. When the lubricating oil moisture content obtained from online inversion reaches or exceeds the alarm threshold, the system automatically outputs corresponding alarm information to remind operators to further inspect, process, or replace the lubricating oil in the main oil tank.

[0045] In one specific implementation, steps S1 to S5 are repeated at preset time intervals to monitor the water content of the lubricating oil in the turbine main oil tank online.

[0046] In this embodiment, the preset time interval can be set according to the turbine's operating conditions, maintenance requirements, and monitoring accuracy requirements. It can be a fixed time interval or adjusted according to changes in the state of the monitored object. By periodically repeating steps S1 to S5, the change in the water content of the turbine's main oil tank lubricating oil can be continuously obtained.

[0047] Specifically, after the system starts, the miniature spectrometer automatically collects the spectral signal of the lubricating oil according to the preset sampling period, and transmits the collected raw spectral data to the data processing module for preprocessing, feature extraction, and model inversion. Finally, it outputs the lubricating oil water content result and its alarm information. The current water content value and its changing trend can also be displayed synchronously on the display interface. The online inversion method is preferably suitable for lubricating oils in the main oil tank of the steam turbine that belong to the same oil type, grade, or base oil system as the modeled sample. This is because different lubricating oils have different base oil compositions, additive systems, and spectral response characteristics. There may be differences between the online monitoring object and the modeling sample in terms of oil type, grade, or formulation system. Therefore, when there are differences between the online monitoring object and the modeling sample in terms of oil type, grade, or formulation system, it is preferable to re-collect the reference spectrum of the corresponding lubricating oil sample and re-establish the water content inversion model that matches the lubricating oil. For lubricating oils of the same type and with basically the same formulation system, the same reference spectrum determination method, water content sensitive band, and water content inversion model can be used. For lubricating oils of different types, it is advisable to establish corresponding reference spectra, preset water content sensitive bands, and water content inversion models separately to improve the accuracy and applicability of the online inversion results.

[0048] In one specific embodiment, the online monitoring system includes a miniature spectrometer, an incident optical fiber, a receiving optical fiber, an anti-pollution probe, a signal acquisition probe, a data processing module, and a result output module. The miniature spectrometer is mounted on the outer side wall of the main oil tank of the steam turbine and connected to the main oil tank shell via an L-shaped bracket. The bracket is equipped with a shock-absorbing structure to reduce the impact of main oil tank operating vibration on spectral acquisition. The detection wavelength range of the miniature spectrometer is 200nm–1000nm, and the spectral resolution is no greater than 2nm. One end of the incident optical fiber is connected to the light source output end of the miniature spectrometer, and the other end extends into the main oil tank. One end of the receiving optical fiber is connected to the signal input end of the miniature spectrometer, and the other end extends into the main oil tank. Both the incident and receiving optical fibers are made of quartz fiber, with a diameter of 0.5mm and a transmission loss of 0.3dB / m. An anti-pollution probe is installed at the end of the incident optical fiber, and a signal acquisition probe is installed at the end of the receiving optical fiber. Both the anti-contamination probe and the signal acquisition probe extend into the main oil tank through pre-drilled mounting holes and are positioned below the minimum operating liquid level to ensure they are submerged in lubricating oil during online monitoring. In this embodiment, the probe is installed 15cm from the bottom of the main oil tank. Both the anti-contamination probe and the signal acquisition probe have titanium alloy shells and sapphire lenses. The data processing module and result output module are housed in an industrial-grade tablet PC. The industrial-grade tablet PC has built-in dedicated analysis software and communicates with the miniature spectrometer to receive raw spectral data, perform preprocessing, call the moisture content inversion model, output moisture content results, and generate alarm information.

[0049] In this embodiment, the online inversion of the water content of the main oil tank lubricating oil of the steam turbine is achieved using the following steps: Step S1: Collect raw spectral data of the lubricating oil. After the system starts, the miniature spectrometer outputs detection light, which is transmitted through an incident optical fiber to the lubricating oil inside the main oil tank. The lubricating oil generates a spectral response to the detection light, which is received by the signal acquisition probe and transmitted back to the miniature spectrometer via a receiving optical fiber, thus obtaining the raw spectral data of the lubricating oil in the main oil tank. The miniature spectrometer automatically performs a spectral acquisition at a preset time interval, which can be set to 5 minutes. To reduce random errors, three sets of raw spectral data are continuously acquired during each acquisition, and the average value is taken as the raw spectral data at that moment.

[0050] Step S2: Preprocess the raw spectral data The raw spectral data obtained in step S1 is input into the data processing module, where noise reduction, baseline correction, and normalization are performed sequentially to obtain the spectral data to be analyzed. Noise reduction eliminates random noise; baseline correction corrects baseline drift caused by light source fluctuations, oil state changes, or probe surface adhesion; and normalization reduces the impact of overall light intensity variations at different sampling times on subsequent feature extraction. These processes improve the comparability of spectral data acquired at different time points, providing a foundation for subsequent feature extraction.

[0051] Step S3: Extract absorbance feature parameters The spectral data obtained in step S2 is analyzed to extract absorbance characteristic parameters within a preset water-sensitive band. In this embodiment, the preset water-sensitive band is determined through pre-experimental calibration, i.e., firstly, standard spectral data of multiple groups of lubricating oil samples with different water contents are collected in the wavelength range of 200nm-1000nm, and then the spectral curves of each group of samples are compared to select the band that responds more significantly to changes in water content as the water-sensitive band. Further, within the determined water-sensitive band, the absorbance peak value, absorbance peak area, and the ratio of absorbance corresponding to different characteristic wavelengths are extracted as absorbance characteristic parameters. The extracted absorbance characteristic parameters can be used individually as input to the water content inversion model, or they can be combined to form a feature vector as input to the water content inversion model.

[0052] Step S4: Establish and call the moisture content inversion model In this embodiment, the moisture content inversion model is established in the following manner: First, multiple sets of lubricating oil samples with different water contents are selected as modeling samples. Using the same miniature spectrometer, optical fiber, and probe structure as in the online monitoring stage, the spectra of each set of lubricating oil samples are acquired to obtain standard spectral data. Then, the same preprocessing operation as in step S2 is performed on each set of standard spectral data, and the corresponding absorbance characteristic parameters are extracted. Finally, the absorbance characteristic parameters of each set of lubricating oil samples are correlated with their known water content standard values ​​to establish a correspondence model between the absorbance characteristic parameters and the water content of the lubricating oil, resulting in a water content inversion model. The known water content standard values ​​are obtained through offline detection. The water content inversion model can be a fitting model or a regression model, as long as it can output the water content of the lubricating oil based on the absorbance characteristic parameters. In the online monitoring stage, the absorbance characteristic parameters obtained in step S3 are input into the water content inversion model to output the water content of the lubricating oil in the main oil tank at the current time.

[0053] Step S5: Output the result and issue an alarm. The lubricating oil water content obtained in step S4 is output to an industrial-grade tablet PC interface and displayed numerically. Simultaneously, the water content results from multiple consecutive time points can be generated into a trend curve to assist operators in observing changes in the lubricating oil water content in the main oil tank. A lubricating oil water content alarm threshold is preset. When the lubricating oil water content obtained in step S4 reaches or exceeds the alarm threshold, the result output module automatically issues an alarm message. The alarm message includes audible and visual alarm messages and / or remote push alarm messages to remind operators to promptly check and address the lubricating oil status in the main oil tank.

[0054] Specifically, steps S1 to S5 are repeated cyclically at preset time intervals to monitor the water content of the lubricating oil in the turbine main oil tank online. After the system starts, the miniature spectrometer automatically collects the spectral signal of the lubricating oil according to the set sampling cycle. The data processing module automatically completes preprocessing, absorbance feature parameter extraction, and water content inversion calculation. The result output module automatically displays the water content result and outputs an alarm message when the threshold is exceeded. Thus, continuous monitoring of the water content of the lubricating oil in the turbine main oil tank can be achieved without frequent manual sampling. The preset water content sensitive band is determined through pre-experimentation. Specifically, within the detection wavelength range of the miniature spectrometer, multiple groups of lubricating oil samples with different water contents are spectrally collected, and the spectral curves of each group of lubricating oil samples are compared and analyzed. The band that shows a significant response to changes in water content and has good distinguishability is selected as the preset water content sensitive band. Furthermore, the preset water-sensitive band can be one band or multiple bands. When multiple bands are used, absorbance characteristic parameters in each band can be extracted and combined to form a feature vector for water content inversion. In an exemplary embodiment, by comparing and analyzing standard spectral data of lubricating oil samples with different water contents in the wavelength range of 200nm-1000nm, the preset water-sensitive band is determined to be 720nm-840nm. Within this band, the absorbance peak value, absorbance peak area, and the ratio of absorbance corresponding to different characteristic wavelengths are extracted as input parameters for the water content inversion model. When establishing the water content inversion model, 30 groups of lubricating oil samples with different water contents are selected as modeling samples. The standard spectral data of each group of lubricating oil samples are obtained using the same spectral acquisition conditions as in the online monitoring stage, and the standard water content values ​​of each group of lubricating oil samples are obtained using conventional offline moisture detection methods. After preprocessing the standard spectral data of each group of lubricating oil samples, absorbance characteristic parameters within the preset water-sensitive band were extracted. These absorbance characteristic parameters were then correlated with the corresponding water content standard values ​​to establish the water content inversion model. Following the establishment of the water content inversion model, 10 additional lubricating oil samples not involved in the modeling were selected as validation samples. The online inversion water content of the validation samples was compared with their offline detection standard values. The validation results show that the established water content inversion model can accurately reflect the trend of lubricating oil water content changes, and the online inversion results have good consistency with the offline detection results, thus verifying the feasibility of using this model for online monitoring of the water content of lubricating oil in the main oil tank of a steam turbine.

[0055] In summary, this embodiment has at least the following technical effects: By collecting the raw spectral data of the lubricating oil in the main oil tank and performing preprocessing, feature extraction, and model inversion, the water content of the lubricating oil can be obtained directly during equipment operation, eliminating the need for frequent manual sampling and offline testing, thereby improving the timeliness of lubricating oil status acquisition. In this invention, each step can be repeated at a preset time interval to continuously monitor the water content of the lubricating oil in the main oil tank of the steam turbine, and output alarm information when the water content exceeds a preset threshold. Therefore, it is beneficial for operators to detect abnormal changes in the water content of the lubricating oil in the main oil tank in a timely manner, and reduce the risk of decreased lubrication performance, oil deterioration or equipment malfunction caused by water entering the lubricating oil. This invention employs a miniature spectrometer located outside the main oil tank of a steam turbine, and completes spectral acquisition in conjunction with a probe located inside the main oil tank via incident and receiving optical fibers. Compared with traditional large-scale field testing equipment, it has the advantages of compact structure, convenient installation, and less impact on the normal operation of the unit, making it more suitable for online applications in the main oil tank scenario. By performing noise reduction, baseline correction, and normalization on the original spectral data, and extracting the absorbance characteristic parameters within the preset water-sensitive band, and then inputting them into the pre-established water content inversion model, the influence of factors such as acquisition noise, light intensity fluctuations, and baseline drift on the analysis results can be reduced, thereby improving the consistency and comparability of the lubricating oil water content inversion results. By installing an anti-contamination probe at the end of the incident optical fiber and a signal acquisition probe at the end of the receiving optical fiber, and placing both within the main oil tank at a position below the minimum normal operating liquid level, the probes are positioned within the lubricating oil coverage area during online monitoring. This facilitates continuous acquisition of the spectral signal of the lubricating oil inside the main oil tank. Furthermore, the anti-contamination probe also reduces the impact of impurities in the oil on the optical fiber ends.

[0056] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for online monitoring and inversion of the water content of lubricating oil in the main oil tank of a steam turbine, characterized in that, Includes the following steps: S1: Collect raw spectral data of the lubricating oil in the main oil tank of the steam turbine; S2: Preprocess the raw spectral data to obtain the spectral data to be analyzed; S3: Extract the absorbance characteristic parameters of the spectral data to be analyzed within a preset water-sensitive band; S4: Input the absorbance characteristic parameters into the pre-established water content inversion model to obtain the water content of the lubricating oil; S5: Output the water content of the lubricating oil, and output an alarm message when the water content of the lubricating oil exceeds a preset threshold.

2. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 1, characterized in that, In step S1, the raw spectral data of the lubricating oil is acquired by a spectrometer, an incident optical fiber, and a receiving optical fiber located outside the main oil tank of the steam turbine. The detection wavelength range of the spectrometer is 200nm to 1000nm, and the spectral resolution of the spectrometer is no greater than 2nm.

3. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 2, characterized in that, An anti-pollution probe is provided at the end of the incident optical fiber, and a signal acquisition probe is provided at the end of the receiving optical fiber.

4. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 3, characterized in that, Both the pollution prevention probe and the signal acquisition probe are installed inside the main oil tank of the steam turbine at a position below the minimum normal operating liquid level.

5. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 1, characterized in that, In step S2, the preprocessing includes noise reduction, baseline correction, and normalization.

6. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 1, characterized in that, In step S3, the absorbance characteristic parameter is the absorbance characteristic quantity of the spectral data to be analyzed in a preset water-sensitive band. The absorbance characteristic parameter includes one or more of the following in the preset water-sensitive band: absorbance peak value, absorbance peak area, ratio of absorbance corresponding to different characteristic wavelengths, first derivative characteristic, and second derivative characteristic.

7. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 1, characterized in that, In step S4, the water content inversion model is a model that establishes the correspondence between absorbance characteristic parameters and water content based on lubricating oil samples with known water content standard values.

8. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 7, characterized in that, The establishment of the moisture content inversion model includes the following steps: Standard spectral data were collected from multiple sets of lubricating oil samples with different water contents; The standard spectral data are preprocessed and the corresponding absorbance characteristic parameters are extracted; as well as The absorbance characteristic parameters are correlated with the known water content standard values ​​of each lubricating oil sample to establish the water content inversion model.

9. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to claim 1, characterized in that, In step S5, the alarm information output is audible and visual alarm information and / or remote push alarm information.

10. The method for online monitoring and inversion of water content in the main oil tank lubricating oil of a steam turbine according to any one of claims 1 to 9, characterized in that: Steps S1 to S5 are repeated at preset time intervals to monitor the water content of the lubricating oil in the turbine main oil tank online.

Images