A device for detecting water content in oil liquid with double probe self-cleaning capability

Abstract

The application discloses a kind of water content detection device in oil liquid with double probe self-cleaning ability, it is related to oil quality detection field, the water content detection device in this oil liquid adopts double probe structure and carries probe self-cleaning device, main control module carries out water content detection based on two different types of probe respectively, and the difference index of two water content prediction values obtained is calculated to judge probe pollution state, when probe pollution risk is larger, discard water content prediction value and trigger probe self-cleaning, until probe state recovers normal, data fusion is carried out to two groups of water content prediction values to obtain final result, this method has the complementary advantage that microwave detection is sensitive to the change of dielectric properties of oil-water mixture and near-infrared detection is sensitive to specific functional groups of water molecules, can effectively improve the detection precision and stability under low water content condition, improve the sensitivity and anti-interference ability of water content detection in oil, suitable for online monitoring of oilfield pipeline, storage tank and refining device.

Description

Technical Field

[0001] This application relates to the field of oil quality testing, and in particular to a device for detecting the water content in oil with dual-probe self-cleaning capability. Background Technology

[0002] In the entire petrochemical industry chain, including oil extraction, refining, storage, and transportation, the detection of water content in oil is crucial. The presence of water can cause numerous problems: in the extraction stage, water increases the difficulty and cost of oil extraction and may combine with acidic components in the oil, causing corrosion to extraction equipment; during refining, water affects the stability and efficiency of oil refining processes, reducing oil quality; during storage and transportation, water can easily cause oil stratification and deterioration, and may freeze at low temperatures, clogging pipelines and other facilities. Accurate detection of water content in oil is key to ensuring the safe and efficient operation of all stages of oil production, processing, storage, and transportation. Currently, there are two common methods for checking the water content in oil: The first method uses near-infrared detection technology, which detects water content by inserting a near-infrared probe into the oil to collect near-infrared spectra. Although near-infrared detection technology is sensitive to the interaction between specific functional groups in oil and water, the composition of oil is extremely complex, containing a variety of hydrocarbon and non-hydrocarbon compounds. When the water content range is wide or the oil composition varies greatly, it is easily affected by other substances, making it difficult to guarantee stable detection accuracy.

[0003] The second method uses microwave detection technology, which involves inserting a microwave probe into the oil to collect microwave signals for water content detection. Microwave detection technology is sensitive to changes in the dielectric properties of oil and can detect water content, but its detection resolution is insufficient for some oil samples with low water content or special components.

[0004] In summary, both near-infrared and microwave detection technologies have significant limitations. Furthermore, during long-term use, the probes are prone to adhering to impurities such as oil, gum, asphalt, and silt from the petroleum, which alters the physical properties of the probe surface (such as dielectric constant), interferes with the detection signal, and further leads to result deviations, misjudgments of the water content in the petroleum, and consequently affects decisions such as adjusting petroleum extraction plans, optimizing refining processes, and ensuring the safety of storage and transportation. Summary of the Invention

[0005] This application addresses the aforementioned problems and technical needs by proposing a device for detecting the water content in oil with dual-probe self-cleaning capability. The technical solution of this application is as follows: A device for detecting the water content in oil with dual-probe self-cleaning capability includes: a microwave probe and a near-infrared probe arranged in the oil to be tested; the microwave probe is connected to a microwave information processing unit, and the near-infrared probe is connected to a near-infrared information processing unit; both the microwave information processing unit and the near-infrared information processing unit are connected to a main control module; a probe self-cleaning device is arranged at each sensor probe, and the two probe self-cleaning devices are respectively connected to the main control module through a self-cleaning drive unit; The microwave information processing unit acquires the microwave signal of the oil to be tested, which shows the change of amplitude and phase with frequency, through a microwave probe and sends it to the main control module. The near-infrared information processing unit acquires the near-infrared spectrum of the oil to be tested through a near-infrared probe and sends it to the main control module. The main control module uses a microwave water content prediction model based on the Bruggenman mixed dielectric constant model to obtain the first predicted water content value from the microwave signal. And the second predicted moisture content value is obtained based on near-infrared spectroscopy using a moisture content neural network prediction model. ; The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference threshold is reached, the two self-cleaning drive units control the self-cleaning devices of their respective connected probes to perform self-cleaning of the corresponding sensor probes and re-acquire microwave signals and near-infrared spectra. The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference threshold is less than the predicted value of the first moisture content, Compared with the second moisture content prediction value Weighted fusion is performed to obtain the water content detection results of the oil to be tested. .

[0006] A further technical solution is that each probe self-cleaning device includes an electromagnetic stirrer, a rotating frame equipped with a permanent magnet, and a cleaning brush. The electromagnetic stirrer is connected to the self-cleaning drive unit, the rotating frame is arranged within the magnetic field range generated by the electromagnetic stirrer, and the cleaning brush is fixed on the rotating frame. The working range of the cleaning brush covers the probe window of the sensor probe. The main control module controls the electromagnetic stirrer to start through the self-cleaning drive unit to generate a magnetic field to drive the rotating frame to rotate the cleaning brush. During the rotation, the cleaning brush contacts the probe window of the corresponding sensor probe to achieve rotational brushing.

[0007] A further technical solution is that each probe self-cleaning device also includes an ultrasonic vibrating plate, which is arranged near the probe window of the corresponding sensor probe and at a predetermined distance from the probe window. The ultrasonic vibrating plate is connected to the self-cleaning drive unit. The main control module controls the ultrasonic vibrating plate to vibrate and decompose the oil stains on the surface of the probe window through the self-cleaning drive unit.

[0008] A further technical solution is that each probe self-cleaning device also includes a microfluidic flushing device, the nozzle of which faces the probe window of the sensor probe, and the microfluidic flushing device is connected to the self-cleaning drive unit; the main control module controls the microfluidic flushing device through the self-cleaning drive unit to spray the built-in cleaning fluid into the probe window of the sensor probe at a predetermined pressure through the nozzle.

[0009] A further technical solution is that each probe self-cleaning device also includes a conductive heating film attached to the probe window of the sensor probe, and the conductive heating film is connected to the self-cleaning drive unit; the main control module controls the conductive heating film to start heating through the self-cleaning drive unit.

[0010] A further technical solution involves the main control module controlling two self-cleaning drive units to control their respective connected probe self-cleaning devices to perform self-cleaning of the corresponding sensor probes, including: The main control module controls the electromagnetic stirrer connected to each self-cleaning drive unit to start so that the cleaning brush can rotate and scrub the probe window surface. After the predetermined scrubbing time is reached, the electromagnetic stirrer is turned off and the ultrasonic vibrating plate is started. After the predetermined vibration time is reached, the ultrasonic vibrating plate is turned off and the microfluidic flushing device is controlled to spray cleaning fluid. At the same time, the conductive heating film is controlled to start heating. After the predetermined spraying time is reached, the microfluidic flushing device and the conductive heating film are turned off.

[0011] The further technical solution is that the main control module calculates the first predicted moisture content value according to the following formula. Compared with the second moisture content prediction value Difference indicators :

[0012] in, It is the absolute difference between two predicted moisture content values ​​within a historical time window. Steady-state standard deviation estimation, It is a parameter for preventing zero loss.

[0013] The further technical solution is to predict the first moisture content. Compared with the second moisture content prediction value Weighted fusion is performed to obtain the water content detection results of the oil to be tested. include: Based on constraint weights according to For the first moisture content prediction Compared with the second moisture content prediction value The moisture content test results were obtained by weighted fusion. ; Alternatively, based on constraint weights according to For the first moisture content prediction Compared with the second moisture content prediction value Residual enhancement fusion was performed to obtain the moisture content detection results. ; in, This represents the residual between two predicted moisture content values ​​and , These are residual enhancement calculation coefficients. , Indicates the confidence parameter and , This represents the absolute difference between two predicted moisture content values. To select the parameters for flow rate, It is the error variance of the microwave moisture content prediction model on the validation dataset. It is the error variance of the moisture content neural network prediction model on the validation dataset. and These are two preset thresholds.

[0014] Its further technical solution is to constrain weights. .

[0015] The further technical solution is that the main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators Whether the difference threshold is reached includes: When the first moisture content prediction value is detected Compared with the second moisture content prediction value Difference indicators Reaching the action threshold The number of consecutive detection cycles reaches the cycle threshold. At that time, determine the difference indicators If the difference threshold is reached, otherwise a difference indicator is determined. The difference threshold was not met; Among them, action threshold The action threshold was calculated using the mean-standard deviation method based on historical data. , The absolute difference between two predicted moisture content values Historical average, The absolute difference between two predicted moisture content values Historical standard deviation This is an empirical coefficient; Or, action threshold The action threshold is determined by using the quantile method on historical data. Take the absolute difference between the two predicted moisture content values The 99th percentile of historical data.

[0016] The beneficial technical effects of this application are: This application discloses an oil water content detection device with dual-probe self-cleaning capability. The device employs a dual-probe structure and incorporates a probe self-cleaning mechanism. First, two different types of probes are used to detect water content, yielding two predicted water content values. A difference index is calculated to determine the probe's contamination status. If the probe has a high risk of contamination, the predicted water content value is discarded, triggering probe self-cleaning. This process continues until the probe returns to normal. Finally, the two sets of predicted water content values ​​are fused to obtain the final result. This method combines the complementary advantages of microwave detection's sensitivity to changes in the dielectric properties of oil-water mixtures with near-infrared detection's sensitivity to specific functional groups of water molecules. It effectively improves detection accuracy and stability under low water content conditions, enhancing the sensitivity and anti-interference capability of oil water content detection. The self-cleaning function also keeps the probes clean, reducing manual maintenance costs and improving the accuracy and stability of oil water content detection. It is suitable for online monitoring of oilfield pipelines, storage tanks, and refining units. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of an oil water content detection device according to an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the working process of the oil water content detection device in one embodiment of this application.

[0019] Figure 3 yes Figure 1 Enlarged view of the probe self-cleaning device at point A in the local structure. Detailed Implementation

[0020] The specific embodiments of this application will be further described below with reference to the accompanying drawings.

[0021] This application discloses a device for detecting the water content in oil with dual-probe self-cleaning capability. Please refer to [link / reference]. Figure 1The oil water content detection device includes two sensor probes arranged in the oil to be tested 100: a microwave probe 1 and a near-infrared probe 2. The probe window of each sensor probe is immersed in the oil to be tested 100. The microwave probe 1 is connected to the microwave information processing unit 3, and the near-infrared probe 2 is connected to the near-infrared information processing unit 4. Both the microwave information processing unit 3 and the near-infrared information processing unit 4 are connected to the main control module 5. A probe self-cleaning device is arranged at each sensor probe, and the two probe self-cleaning devices are respectively connected to the main control module 5 through a self-cleaning drive unit 6. The main control module 5, the two microwave information processing units, and the two self-cleaning drive units 6 can be different software control modules in the same chip, or they can be implemented by combining multiple chips, which will not be elaborated here.

[0022] To ensure sufficient contact between the probe windows of the two sensor probes and the oil to be tested 100, in another embodiment, the water content detection device for the oil further includes a sampling tank 8. The sampling tank 8 employs a flow-through pipe structure, and the oil to be tested 100 flows through the sampling tank 8 at a constant flow rate. The two sensor probes are positioned upstream and downstream of the oil to be tested 100, respectively, along the flow direction of the oil to be tested, with the probe windows of both sensor probes immersed in the oil to be tested 100. The oil to be tested 100 in this application includes crude oil, transformer oil, etc., and the sampling tank 8 is made of oil-resistant and light-transmitting polytetrafluoroethylene (PTFE).

[0023] The working process of this oil water content detection device is as follows, please refer to it. Figure 2 The flowchart shown: Step 210: The microwave information processing unit 3 collects the microwave signal of the oil to be tested 100, whose amplitude and phase change with frequency through the microwave probe 1 and sends it to the main control module 5. The near-infrared information processing unit 4 collects the near-infrared spectrum of the oil to be tested through the near-infrared probe 2 and sends it to the main control module 5.

[0024] Two sensor units acquire data synchronously. Microwave probe 1 acquires microwave signals whose amplitude and phase vary with frequency, while near-infrared probe 2 acquires spectral signals within a specific wavelength range. In one embodiment, microwave probe 1 employs a microstrip antenna structure, operating at frequencies ranging from 5 GHz to 20 GHz. Microwave probe 1 acquires microwave signals whose amplitude and phase vary with frequency. In one embodiment, near-infrared probe 2 is equipped with an optical filtering system covering a wavelength range of 1200 nm to 2500 nm, combined with a fiber optic spectrometer to acquire the near-infrared transmission or diffuse reflectance spectral signals of the oil 100 to be tested.

[0025] Step 220: The main control module 5 uses the microwave water content prediction model based on the Bruggenman mixed dielectric constant model to obtain the first predicted water content value based on the microwave signal. And the second predicted moisture content value is obtained based on near-infrared spectroscopy using a moisture content neural network prediction model. The details are as follows: (1) Obtain the first predicted moisture content value

[0026] The microwave moisture content prediction model in this application utilizes the differences in electromagnetic properties such as wave velocity and phase shift when microwaves propagate in media with different dielectric properties to detect moisture content. Microwaves interact with the equivalent dielectric constant of the medium, and their propagation wave velocity and phase shift are quantitatively correlated with the moisture content of the medium. To achieve accurate extraction of dielectric characteristics and quantitative analysis of moisture content, the derivation and modeling of parameters such as microwave propagation wave velocity and phase shift are performed using the following method: A microwave moisture content analyzer detects moisture content by utilizing the principle that microwaves travel at different speeds in different media. The formula for the speed of microwave transmission is known to be: (1) in, It is the wavenumber of microwave transmission and the unit is m / s. It is the wave speed of microwaves propagating under vacuum conditions, and the unit is m / s. It is the equivalent dielectric constant of the medium through which microwaves pass, and its unit is F / m. The magnetic permeability of the medium material is expressed in H / m. Oil is a non-magnetic medium, and the magnetic permeability of its medium material can generally be directly equivalent to 1.

[0027] Since the wave velocity of microwaves is difficult to measure directly in practice, it is converted into an indirect measurement by detecting the phase shift that occurs when the microwave signal passes through the same transmission line length. The phase shift formula is as follows: (2) in, The value represents the phase shift of the microwave, and the unit is rad. The distance traversed through a medium containing a water mixture is expressed in meters (m). It is the phase shift constant and its unit is rad / m.

[0028] When the transmission distance is fixed, the phase shift constant The larger the value, the greater the phase shift of the microwave, and the greater the phase shift constant of the microwave. The formula is: (3) In the formula, The angular frequency of the microwave signal is expressed in rad / s.

[0029] Substituting equations (3) and (2) into equation (1) yields the phase shift value. With equivalent dielectric constant The expression is: (4) The phase shift value after microwave passes through the oil under test is measured. After determining the size, the equivalent dielectric constant of the oil to be tested, 100, can be calculated using formula (4). The equivalent dielectric constant of the oil to be tested, 100. The water content in the tested oil 100 is related to the ratio of oil to water, and its equivalent dielectric constant is related to the water content in the tested oil 100. The relationship is expressed using the Bruggenman mixed dielectric constant model as follows: (5) in, It is the dielectric constant of pure oil, and the unit is F / m. It is the dielectric constant of pure water, and its unit is F / m.

[0030] By acquiring oil samples with water content varying from 0% to 100% at predetermined step sizes, and collecting the microwave signal-water content and phase shift-water content relationships for each oil sample, a microwave water content prediction model can be established based on the Bruggenman mixed dielectric constant model. This microwave water content prediction model characterizes the relationship between microwave signal and water content. Therefore, using this microwave water content prediction model, a first predicted water content value can be obtained based on the collected microwave signal. .

[0031] (2) Obtain the second predicted moisture content value

[0032] Near-infrared moisture content analysis detects moisture content by utilizing the differences in the molecular vibrational absorption characteristics of water molecules to near-infrared light at different moisture contents. The main control module first performs preprocessing operations on the near-infrared spectrum, including wavelet transform denoising and Savitzky-Golay convolution smoothing. Assuming the near-infrared spectrum after preprocessing is , They are respectively The near-infrared spectrum contains several spectral data points, each containing absorbance values ​​at multiple wavelengths. The preprocessed near-infrared spectrum is then input into a pre-trained CNN-based neural network prediction model for water content, yielding a second predicted water content value. .

[0033] Step 230: The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators Whether the difference threshold has been reached.

[0034] In one embodiment, the main control module calculates the first moisture content prediction value according to the following formula. Compared with the second moisture content prediction value Difference indicators : (6) in, It is the absolute difference between two predicted moisture content values ​​within a historical time window. Steady-state standard deviation estimation, It is a parameter for preventing zero loss.

[0035] Then the difference indicators With action threshold Comparison, when the first predicted moisture content is detected Compared with the second moisture content prediction value Difference indicators Reaching the action threshold The number of consecutive detection cycles reaches the cycle threshold. At that time, determine the difference indicators If the difference threshold is reached, otherwise a difference indicator is determined. The difference threshold was not reached. The periodic threshold here... For custom values, such as those that can generally be set to... It consists of 70 testing cycles.

[0036] To ensure the accuracy of the results, the action threshold... There are two methods for determining this based on historical data: (1) It is calculated by using the mean-standard deviation method on historical data. Specifically: Action threshold , The absolute difference between two predicted moisture content values Historical average, The absolute difference between two predicted moisture content values The historical standard deviation. This is an empirical coefficient, usually taken as 3 or 4.

[0037] (2) Determined by using the quantile method on historical data, specifically: the action threshold is determined. Take the absolute difference between the two predicted moisture content values The 99th percentile of historical data.

[0038] In addition, an early warning function can be set up to detect the first predicted moisture content. Compared with the second moisture content prediction value Difference indicators Reaching the warning threshold The number of consecutive detection cycles reaches the cycle threshold. At that time, a pollution warning is triggered, indicating a potential risk of probe contamination. For example, it is generally possible to take For 30 detection cycles. Warning threshold. The warning threshold can also be determined based on historical data using the mean-standard deviation method or the quantile method. When using the mean-standard deviation method, empirical coefficient Take less than The value can be, for example, 1 or 2. When using the quantile method, the absolute difference between two predicted moisture content values ​​can be taken. The 95th percentile of historical data is used as the warning threshold. .

[0039] Step 240, the main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference threshold is reached, it indicates that the two moisture content prediction values ​​differ significantly. This is often caused by the sensor probe being contaminated by oil, resulting in abnormal data acquisition. At this time, the accuracy and reliability of the two moisture content prediction values ​​are low, and the conditions for data fusion are not met. Therefore, the main control module controls the self-cleaning device of the probe connected to each of the two self-cleaning drive units 6 to perform self-cleaning on the corresponding sensor probe and re-acquire microwave signals and near-infrared spectra.

[0040] The two probe self-cleaning devices in this application have the same structure and relative arrangement with the corresponding sensor probes, only differing in that... Figure 1 Taking the probe self-cleaning device arranged at microwave probe 1 as an example... Figure 1 An enlarged view of structure A, shown by the dashed line, is as follows: Figure 3 As shown.

[0041] Each probe self-cleaning device includes an electromagnetic stirrer 71, a rotating frame 72 equipped with a permanent magnet, and a cleaning brush 73. The electromagnetic stirrer 71 is electrically connected to the self-cleaning drive unit 6. The rotating frame 72 is arranged within the magnetic field generated by the electromagnetic stirrer 71, typically at the center of the magnetic field. The cleaning brush 73 is fixed to the rotating frame 72. To improve cleaning efficiency, multiple cleaning brushes 73 are arranged at different positions around the circumference of the rotating frame 72, such as... Figure 3Taking the arrangement of two cleaning brushes as an example, the cleaning brush 73 is generally a silicone brush. The working range of the cleaning brush 73 covers the probe window of the sensor probe. The main control module 5 controls the electromagnetic stirrer 71 to start through the self-cleaning drive unit 6 to generate a magnetic field to drive the rotating frame 72 to rotate the cleaning brush 73. During the rotation, the cleaning brush 73 contacts the probe window of the corresponding sensor probe 1 to achieve rotational brushing, thereby wiping away the oil and large particles attached to the surface of the probe window.

[0042] In another further embodiment, each probe self-cleaning device also includes an ultrasonic vibrating plate 74, which is arranged near the probe window of the corresponding sensor probe 1 and spaced at a predetermined distance from the probe window, thereby avoiding direct contact with the sensor probe. In practice, the ultrasonic vibrating plate 74 is typically arranged on the wall of the oil pipe near the probe window, or... Figure 1 On the wall near the probe window of sampling cell 8, Figure 3 The location of the ultrasonic vibrating plate 74 is shown only as an example. Figure 1 The ultrasonic vibrating element 74 is not shown in detail in structure A. The ultrasonic vibrating element 74 is electrically connected to the self-cleaning drive unit 6. The main control module 5 controls the vibration of the ultrasonic vibrating element 74 through the self-cleaning drive unit 6 to decompose the oil stains on the surface of the probe window.

[0043] In another further embodiment, each probe self-cleaning device also includes a microfluidic flushing device 75, the nozzle 75a of which faces the probe window of the sensor probe 1. The microfluidic flushing device 75 is electrically connected to the self-cleaning drive unit 6. The main control module 5 controls the microfluidic flushing device 75 to spray the built-in cleaning fluid onto the probe window of the sensor probe 1 through the nozzle 75a at a predetermined pressure via the self-cleaning drive unit 6. In one embodiment, the spray pressure of the microfluidic flushing device 75 is 0.1 MPa to 0.3 MPa, and the cleaning fluid used is, for example, anhydrous ethanol. When the two sensor probes in the oil content detection device are placed in the oil to be tested 100 with a constant flow rate, the microfluidic flushing device 75 is located upstream of the sensor probes.

[0044] In another further embodiment, each probe self-cleaning device also includes a conductive heating film 76 attached to the probe window of the sensor probe, the conductive heating film 76 being electrically connected to the self-cleaning drive unit 6. The main control module 5 controls the conductive heating film 76 to start heating via the self-cleaning drive unit 6. In one embodiment, the conductive heating film 76 provides a heating effect of 50°C to 80°C.

[0045] When the probe self-cleaning device includes all the above components, the main control module controls the respective connected probe self-cleaning devices to perform self-cleaning of the corresponding sensor probes through two self-cleaning drive units. The process includes: (1) Control the start of the respective connected electromagnetic stirrers 71 to use the cleaning brush 73 to rotate and brush the probe window surface for a predetermined time, thereby wiping away the oil stains attached to the probe window surface and ensuring that larger deposits on the probe window surface are removed. The predetermined brushing time can be customized, for example, set to 15 seconds.

[0046] (2) After the predetermined scrubbing time is reached, turn off the electromagnetic stirrer 71, start the ultrasonic vibrator 74, and continue vibrating for the predetermined time. The ultrasonic vibrator 74 vibrates at a predetermined frequency, transmitting the vibration through the liquid and generating a cavitation effect, which can decompose the fine oil stains on the probe window surface. In one example, the operating frequency of the ultrasonic vibrator 74 is 20kHz~50kHz. The predetermined vibration time can be customized, for example, set to 10 seconds.

[0047] (3) After reaching the predetermined vibration duration, the ultrasonic vibration plate 74 is turned off, the microfluidic flushing device 75 is controlled to spray cleaning fluid, and the conductive heating film 76 is simultaneously activated to heat. After the predetermined spraying duration, the microfluidic flushing device 75 and the conductive heating film 76 are turned off. The microfluidic flushing device 75 continuously flushes the probe window by spraying cleaning fluid onto it, ensuring that the cleaning fluid can remove small particles of oil decomposed by ultrasound and dissolve the oil on the probe surface. The conductive heating film 76 is activated simultaneously with the spraying of cleaning fluid, heating the area around the probe to accelerate the dissolution of oil and prevent the oil from re-condensing at low temperatures. The predetermined spraying duration can be customized, for example, set to 10 seconds.

[0048] After completing the above cleaning process, the main control module returns to step 210 to collect microwave signals and near-infrared spectra and detect the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference index If the difference threshold is still reached, continue to the next cleaning cycle to clean again according to the above process; otherwise, proceed to step 250 below.

[0049] Step 250, the main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference is less than the difference threshold, it indicates that the two moisture content predictions are quite similar. In this case, it means that the risk of contamination of the two sensor probes is low, and the first moisture content prediction value is further adjusted. Compared with the second moisture content prediction value Weighted fusion was performed to obtain the water content test results for the oil to be tested (100%). .

[0050] When performing data fusion, one approach is based on constraint weights. according to For the first moisture content prediction Compared with the second moisture content prediction value Weighted fusion is performed to obtain the moisture content test results. .

[0051] Another approach is based on constraint weights. according to For the first moisture content prediction Compared with the second moisture content prediction value Residual enhancement fusion was performed to obtain the moisture content detection results. .in, This represents the residual between two predicted moisture content values, and: (7) In the above formula, This is the residual enhancement calculation coefficient, taken as 0.5 or 0.7. Intermediate function. : (8) in, Indicates the confidence parameter and , This represents the absolute difference between two predicted moisture content values. To select the parameters for flow rate. It is the error variance of the microwave moisture content prediction model on the validation dataset. It is the error variance of the moisture content neural network prediction model on the validation dataset. and These are two preset thresholds. Typically, 10% of the measurement range is used. Typically, 30% of the measurement range is used.

[0052] Compared to ordinary weighted fusion, residual-enhanced fusion can further improve the fusion effect, thus obtaining more accurate and reliable moisture content detection results. .

[0053] Regardless of whether a standard weighted fusion method or a residual-enhanced fusion method is used, constraint weights are required. To further improve the fusion effect, this constraint weight Instead of directly taking experience points, it follows... Calculated.

[0054] This device is suitable for online water content detection in complex oil systems. It can be further equipped with a display and alarm module for on-site data output. The main control module can also be further connected to a host computer for communication, supporting MODBUS, Ethernet or wireless communication. It can be embedded into an embedded processing system for online oil detection to realize remote monitoring of water in oil and issue an alarm when the water content exceeds the limit.

[0055] The above descriptions are merely preferred embodiments of this application, and this application is not limited to the above embodiments. It is understood that other improvements and variations that can be directly derived or conceived by those skilled in the art without departing from the spirit and concept of this application should be considered to be included within the protection scope of this application.

Claims

1. A device for detecting water content in oil with dual-probe self-cleaning capability, characterized in that, The oil water content detection device includes: a microwave probe and a near-infrared probe arranged in the oil to be tested. The microwave probe is connected to a microwave information processing unit, and the near-infrared probe is connected to a near-infrared information processing unit. Both the microwave information processing unit and the near-infrared information processing unit are connected to the main control module. A probe self-cleaning device is arranged at each sensor probe, and the two probe self-cleaning devices are respectively connected to the main control module through a self-cleaning drive unit. The microwave information processing unit acquires the microwave signal of the oil to be tested, which shows the change of amplitude and phase with frequency, through a microwave probe and sends it to the main control module. The near-infrared information processing unit acquires the near-infrared spectrum of the oil to be tested through a near-infrared probe and sends it to the main control module. The main control module uses a microwave water content prediction model based on the Bruggenman mixed dielectric constant model to obtain the first predicted water content value from the microwave signal. And the second predicted moisture content value is obtained based on near-infrared spectroscopy using a moisture content neural network prediction model. ; The main control module calculates the first predicted moisture content using the following formula. Compared with the second moisture content prediction value Difference indicators : in, It is the absolute difference between two predicted moisture content values ​​within a historical time window. Steady-state standard deviation estimation, These are parameters for preventing zero-level protection. The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference threshold is reached, the two self-cleaning drive units control the self-cleaning devices of their respective connected probes to perform self-cleaning of the corresponding sensor probes and re-acquire microwave signals and near-infrared spectra. The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators When the difference threshold is less than the predicted value of the first moisture content, Compared with the second moisture content prediction value Weighted fusion is performed to obtain the water content detection results of the oil to be tested. .

2. The oil water content detection device according to claim 1, characterized in that, Each probe self-cleaning device includes an electromagnetic stirrer, a rotating frame equipped with a permanent magnet, and a cleaning brush. The electromagnetic stirrer is connected to the self-cleaning drive unit, the rotating frame is arranged within the magnetic field range generated by the electromagnetic stirrer, and the cleaning brush is fixed on the rotating frame. The working range of the cleaning brush covers the probe window of the sensor probe. The main control module controls the electromagnetic stirrer to start through the self-cleaning drive unit to generate a magnetic field to drive the rotating frame to rotate the cleaning brush. During the rotation, the cleaning brush contacts the probe window of the corresponding sensor probe to achieve rotational cleaning.

3. The oil water content detection device according to claim 2, characterized in that, Each probe self-cleaning device also includes an ultrasonic vibrating plate, which is arranged near the probe window of the corresponding sensor probe and at a predetermined distance from the probe window. The ultrasonic vibrating plate is connected to the self-cleaning drive unit. The main control module controls the ultrasonic vibrating plate to vibrate and decompose the oil stains on the surface of the probe window through the self-cleaning drive unit.

4. The oil water content detection device according to claim 3, characterized in that, Each probe self-cleaning device also includes a microfluidic flushing device. The nozzle of the microfluidic flushing device faces the probe window of the sensor probe. The microfluidic flushing device is connected to the self-cleaning drive unit. The main control module controls the microfluidic flushing device through the self-cleaning drive unit to spray the built-in cleaning fluid into the probe window of the sensor probe at a predetermined pressure through the nozzle.

5. The oil water content detection device according to claim 4, characterized in that, Each probe self-cleaning device also includes a conductive heating film attached to the probe window of the sensor probe, and the conductive heating film is connected to the self-cleaning drive unit; the main control module controls the conductive heating film to start heating through the self-cleaning drive unit.

6. The oil water content detection device according to claim 5, characterized in that, The main control module controls the self-cleaning devices of the corresponding sensor probes through two self-cleaning drive units to perform self-cleaning, including: The main control module controls the electromagnetic stirrer connected to each self-cleaning drive unit to start so that the cleaning brush can rotate and scrub the probe window surface. After the predetermined scrubbing time is reached, the electromagnetic stirrer is turned off and the ultrasonic vibrating plate is started. After the predetermined vibration time is reached, the ultrasonic vibrating plate is turned off and the microfluidic flushing device is controlled to spray cleaning fluid. At the same time, the conductive heating film is controlled to start heating. After the predetermined spraying time is reached, the microfluidic flushing device and the conductive heating film are turned off.

7. The oil water content detection device according to claim 1, characterized in that, For the first moisture content prediction Compared with the second moisture content prediction value Weighted fusion is performed to obtain the water content detection results of the oil to be tested. include: Based on constraint weights according to For the first moisture content prediction Compared with the second moisture content prediction value The moisture content test results were obtained by weighted fusion. ; Alternatively, based on constraint weights according to For the first moisture content prediction Compared with the second moisture content prediction value Residual enhancement fusion was performed to obtain the moisture content detection results. ; in, This represents the residual between two predicted moisture content values ​​and , These are residual enhancement calculation coefficients. , Indicates the confidence parameter and , This represents the absolute difference between two predicted moisture content values. To select the parameters for flow rate, It is the error variance of the microwave moisture content prediction model on the validation dataset. It is the error variance of the moisture content neural network prediction model on the validation dataset. and These are two preset thresholds.

8. The oil water content detection device according to claim 7, characterized in that, Constraint weights .

9. The oil water content detection device according to claim 1, characterized in that, The main control module detects the first predicted moisture content value. Compared with the second moisture content prediction value Difference indicators Whether the difference threshold is reached includes: When the first moisture content prediction value is detected Compared with the second moisture content prediction value Difference indicators Reaching the action threshold The number of consecutive detection cycles reaches the cycle threshold. At that time, determine the difference indicators If the difference threshold is reached, otherwise a difference indicator is determined. The difference threshold was not met; Among them, action threshold The action threshold was calculated using the mean-standard deviation method based on historical data. , The absolute difference between two predicted moisture content values Historical average, The absolute difference between two predicted moisture content values Historical standard deviation This is an empirical coefficient; Or, action threshold The action threshold is determined by using the quantile method on historical data. Take the absolute difference between the two predicted moisture content values The 99th percentile of historical data.

Images