Method and device for detecting water content of liquid sample based on energy absorption sensor

Abstract

The present application relates to the field of liquid sample detection, and specifically provides a liquid sample water content detection method and device based on an energy absorption sensor. The method comprises: preprocessing multi-modal original radio frequency signals collected by the energy absorption sensor to obtain target filtered signals; converting the target filtered signals of different modalities into target corrected signals according to a preset error compensation model; wherein the preset error compensation model at least includes one of the following: flow rate-response lag compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation; performing weighted summation on the target corrected signals of different modalities to obtain a fusion target signal; determining the average signal intensity decay rate in each sampling period based on the fusion target signal to obtain an average signal intensity decay rate sequence, and converting it into a liquid sample water content sequence according to a preset conversion relationship. The technical scheme provided by the present application improves the accuracy of liquid sample water content detection.

Description

Technical Field

[0001] This invention relates to the field of liquid sample detection technology, and in particular to a method and apparatus for detecting the water content of liquid samples based on an energy absorption sensor. Background Technology

[0002] In many fields such as oil extraction, chemical production, and food processing, the water content of liquid samples is a crucial testing indicator, and its accuracy directly affects production efficiency, product quality, and safety control. Traditional methods for detecting the water content of liquid samples include distillation, Karl Fischer titration, capacitance, and microwave methods, but these methods generally have certain limitations. For example, distillation is cumbersome and has a long testing cycle, making it difficult to meet the needs of real-time online testing; capacitance is easily affected by environmental factors such as liquid sample temperature and pressure, resulting in unstable detection accuracy; and microwave methods are susceptible to signal interference when an emulsion layer is present in the liquid sample, leading to significant deviations in the test results.

[0003] Energy absorption sensors, which detect energy decay characteristics of radio frequency signals in liquid samples with different water contents, have gained increasing attention due to their advantages such as fast response speed and online detection capability. However, in practical applications, directly analyzing the raw radio frequency energy decay signal acquired by the energy absorption sensor leads to insufficient detection accuracy. Summary of the Invention

[0004] The present invention provides a liquid sample moisture content detection method, device, electronic device, storage medium, and computer program product based on an energy absorption sensor, which improves the accuracy of liquid sample moisture content detection to a certain extent.

[0005] In a first aspect, the present invention provides a method for detecting the water content of a liquid sample based on an energy absorption sensor, the method comprising:

[0006] The raw radio frequency signals of different modes acquired by the energy absorption sensor are preprocessed to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point;

[0007] According to the preset error compensation model, the target filtered signals of different modes are converted into target correction signals respectively; wherein, the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation.

[0008] The target correction signals of different modalities are weighted and summed to obtain the fused target signal; wherein the sum of the weights corresponding to the target correction signals of different modalities is 1.

[0009] Based on the fused target signal, the average signal strength attenuation rate in each sampling period is determined according to the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, thus obtaining the average signal strength attenuation rate sequence.

[0010] Based on a preset conversion relationship, the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence is converted into liquid sample water content to obtain a liquid sample water content sequence.

[0011] In one embodiment of the present invention, the raw radio frequency signals of different modes acquired by the energy absorption sensor are preprocessed to obtain the target filtered signal, including:

[0012] The raw radio frequency signals of different modes collected are cleaned to obtain standard radio frequency signals for each mode; wherein, the data cleaning includes filling missing values, handling outliers, and normalizing environmental parameters of the raw radio frequency signals.

[0013] The standard radio frequency signal is subjected to multi-layer filtering to obtain the target filtered signal.

[0014] In one embodiment of the present invention, the signals acquired by the energy absorption sensor further include temperature signals and pressure signals. The raw radio frequency signals of different modes are cleaned to obtain standard radio frequency signals for each mode, including:

[0015] If the number of consecutive missing data in the original radio frequency signal is less than or equal to the preset number of samplings, the missing region is linearly interpolated using the value adjacent to the missing region to obtain the initial repair signal.

[0016] If the number of consecutive missing data in the original radio frequency signal is greater than the preset number of samplings, the initial repair signal is predicted and filled using a pre-trained long short-term memory network model based on the changing trend of historical data corresponding to the missing region and the initial repair signal to obtain the first radio frequency attenuation signal.

[0017] Calculate the average value and standard deviation of the first radio frequency attenuation signal;

[0018] An outlier is defined as a value where the absolute value of the difference between the value in the first radio frequency attenuation signal and the average value is greater than a preset standard deviation, or a value in the first radio frequency attenuation signal is greater than the limit value of liquid sample energy absorption; wherein, the preset standard deviation is the product of the standard deviation and a first preset multiple.

[0019] The abnormal point corresponding to the abnormal value is taken as the center of the first preset window, and the median of the values ​​in the first preset window is determined as the value of the abnormal point to obtain the second radio frequency attenuation signal.

[0020] Based on the temperature signal and the pressure signal, the energy attenuation values ​​of each sampling point in the second radio frequency attenuation signal are converted into target energy attenuation values ​​under standard conditions to obtain a standard radio frequency signal.

[0021] In one embodiment of the present invention, the standard radio frequency signal is subjected to multi-layer filtering to obtain a target filtered signal, including:

[0022] In the standard radio frequency signal, the center value of the second preset window is replaced with the median of the second preset window to obtain the first filtered signal; wherein, the size of the second preset window when the liquid sample is in a stable period is larger than the size of the second preset window when the liquid sample is in a sudden change period;

[0023] The first filtered signal is decomposed into wavelet packets to obtain the second filtered signal;

[0024] The second filtered signal is smoothed using the Kalman filter algorithm to obtain the target filtered signal.

[0025] In one embodiment of the present invention, the target filtered signals of different modes are converted into target correction signals according to a preset error compensation model, including:

[0026] Real-time acquisition of liquid sample flow rate data; calculation of signal delay time corresponding to the liquid sample flow rate data based on a preset mapping relationship model; and shifting the target filtered signal at the current moment forward based on the signal delay time to obtain a time-series target correction signal; wherein the sampling frequency of the liquid sample flow rate data is the same as the sampling frequency of the original radio frequency signal;

[0027] The time-series target correction signal is subjected to frequency domain transformation to identify the bubble signal amplitude; and the time-series target correction signal is subjected to wavelet decomposition, and the low-frequency component representing the effective signal and the high-frequency component representing the bubble interference are determined based on the bubble signal amplitude; the part of the high-frequency component whose amplitude exceeds the preset amplitude is set to 0, and the bubble target correction signal is obtained by inverse wavelet transformation;

[0028] Based on the thickness of the emulsion layer in the liquid sample and the emulsion layer thickness correction factor, the bubble target correction signal is corrected to obtain the emulsion layer interference correction signal;

[0029] Based on a preset system error model, the emulsion layer interference correction signal is converted into a target correction signal; the preset system error compensation model is used to calibrate the error of the energy absorption sensor; the preset system error compensation model is a linear transformation model constructed by taking the radio frequency signal of the standard liquid sample as the dependent variable and the radio frequency signal measured by the energy sensor on the standard liquid sample as the independent variable.

[0030] In one embodiment of the present invention, after the step of weighted summation of the target correction signals of different modalities to obtain the fused target signal, the method further includes:

[0031] The target signal at the current moment is weighted and summed with the target signal at the previous moment using a second preset weighting coefficient to obtain the target stable signal; wherein the sum of the weights of the target signal at the current moment and the target signal at the previous moment is 1; wherein the weighting coefficient of the target signal at the current moment when the liquid sample is in a stable state is less than the weighting coefficient of the target signal at the current moment when the liquid sample is in an unstable state.

[0032] Accordingly, the step of determining the average signal strength attenuation rate in each sampling period based on the fused target signal and the number of samples in the horizontal and vertical polarization directions within each sampling period, and obtaining the average signal strength attenuation rate sequence, includes:

[0033] Based on the target stable signal, the average signal strength attenuation rate within the current sampling period is determined according to the number of samples taken in the horizontal and vertical polarization directions within the current sampling period, thus obtaining an average signal strength attenuation rate sequence.

[0034] In one embodiment of the present invention, based on the fused target signal, the average signal strength attenuation rate in each sampling period is determined according to the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, resulting in an average signal strength attenuation rate sequence, including:

[0035] For each sampling period, the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction are calculated based on the target fused signal in the current sampling period; or, the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction are calculated based on the target stable signal in the current sampling period.

[0036] The horizontal polarizability and vertical polarizability are determined based on the number of samples taken in the horizontal polarization direction and the number of samples taken in the vertical polarization direction within the current sampling period.

[0037] The average signal strength attenuation rate of the current sampling period is obtained by adding the product of the horizontal polarization rate and the average level signal strength attenuation rate, and the product of the vertical polarization rate and the average vertical signal strength attenuation rate.

[0038] The average signal strength attenuation rate of each sampling period is spliced ​​together to obtain the average signal strength attenuation rate sequence.

[0039] In one embodiment of the present invention, after the step of converting the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence into liquid sample water content based on a preset conversion relationship to obtain a liquid sample water content sequence, the method further includes:

[0040] Calculate the sub-mean and sub-standard deviation of the liquid sample water content in each sampling period within the third preset window in the liquid sample water content sequence;

[0041] If the absolute value of the difference between the liquid sample moisture content and the sub-mean value in the moisture content sequence for m consecutive sampling periods is greater than the preset sub-standard deviation, the moisture content of the current sampling period is replaced with the average moisture content of the previous p sampling periods of the current sampling point; wherein, the preset sub-standard deviation is the product of the sub-standard deviation and the second preset multiple;

[0042] If the absolute value of the difference between the liquid sample water content and the sub-mean value at n consecutive sampling points in the water content sequence is greater than the preset sub-standard deviation, the health diagnosis of the energy absorption sensor is triggered; where n > m.

[0043] In one embodiment of the present invention, the energy absorption sensor is an array sensor probe, which includes multiple sub-energy absorption sensors, including normal region probes and abnormal region probes; each sub-energy absorption sensor is configured to acquire raw radio frequency signals of different modes; after preprocessing the raw radio frequency signals of different modes acquired by the energy absorption sensors to obtain the target filtered signal, the method further includes:

[0044] For each mode of the target filtered signal in each of the sub-energy absorption sensors, calculate the deviation rate between the target filtered signal and the array average signal corresponding to the target filtered signal;

[0045] If the deviation rate of k consecutive sampling points is greater than or equal to a preset deviation, the sub-energy absorption sensor corresponding to the target filtered signal is determined to be an abnormal region probe.

[0046] Different weighting coefficients are assigned to the normal region probe and the abnormal region probe respectively, and spatial fusion is performed in the same dimension to obtain the target fusion filtered signal.

[0047] Secondly, the present invention provides a liquid sample moisture content detection device based on an energy absorption sensor, the liquid sample moisture content detection device based on an energy absorption sensor comprising:

[0048] The signal preprocessing module is used to preprocess the raw radio frequency signals of different modes acquired by the energy absorption sensor to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point;

[0049] An error correction module is used to convert target filtered signals of different modes into target corrected signals according to a preset error compensation model; wherein the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation.

[0050] The signal fusion module is used to perform weighted summation of target correction signals of different modalities to obtain a fused target signal; wherein the sum of the weights corresponding to the target correction signals of different dimensions is 1;

[0051] The signal strength attenuation rate calculation module is used to determine the average signal strength attenuation rate in each sampling period based on the fused target signal and the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, so as to obtain the average signal strength attenuation rate sequence.

[0052] The moisture content conversion module is used to convert the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence into the liquid sample moisture content based on a preset conversion relationship, so as to obtain the liquid sample moisture content sequence.

[0053] Thirdly, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the liquid sample moisture content detection method based on the energy absorption sensor described above.

[0054] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the liquid sample moisture content detection method based on an energy absorption sensor as described in any of the preceding claims.

[0055] Fifthly, the present invention provides a computer program product, including a computer program, which, when executed by a processor, causes the computer to perform the liquid sample moisture content detection method based on an energy absorption sensor as described in any of the preceding claims.

[0056] This invention provides a method for detecting the water content of liquid samples based on an energy absorption sensor. The method preprocesses the raw radio frequency signals acquired by the energy absorption sensor operating in different frequency bands to obtain target filtered signals of different modes. Then, according to a preset error compensation model, the target filtered signals of different modes are corrected to obtain a target corrected signal. The target corrected signals of different modes are then weighted and summed to obtain a fused target signal. Based on the fused target signal, the average signal intensity attenuation rate in each sampling period is determined according to the number of samples taken in the horizontal and vertical polarization directions within each sampling period, resulting in an average signal intensity attenuation rate sequence. Finally, based on a preset conversion relationship, the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence can be converted into the water content of the liquid sample to obtain a water content sequence, thereby improving the accuracy of liquid sample water content detection to a certain extent. Attached Figure Description

[0057] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other embodiments based on these drawings without creative effort.

[0058] Figure 1 This is a flowchart illustrating a liquid sample moisture content detection method based on an energy absorption sensor, according to an embodiment of the present invention.

[0059] Figure 2 This is a schematic diagram of the iterative process of liquid sample water content factor based on an energy absorption sensor, provided in an embodiment of the present invention.

[0060] Figure 3 This is a schematic diagram of a liquid sample moisture content detection device based on an energy absorption sensor, provided in an embodiment of the present invention.

[0061] Figure 4 This is a schematic diagram of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0062] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0063] In offshore oil production and oil processing, it is often necessary to detect the water content in oil. This is primarily achieved using energy absorption sensors (EAS), which detect water content by leveraging the energy attenuation characteristics of radio frequency (RF) signals in liquid samples with varying water contents. EAS offers advantages such as fast response and online detection, and has gained increasing attention. However, the raw RF signals acquired by EAS are susceptible to factors such as missing values, outliers, environmental parameter fluctuations, sensor calibration errors, and emulsion layers in the liquid sample, leading to insufficient detection accuracy. Furthermore, the detection range of single-mode RF signals is limited, making it difficult to adapt to the diverse liquid sample detection requirements under different operating conditions. Therefore, it is crucial to effectively preprocess the raw RF signals, compensate for various errors, and fuse multi-mode signals to improve the accuracy of liquid sample water content detection.

[0064] Based on this, the present invention provides a liquid sample moisture content detection method based on an energy absorption sensor. The method preprocesses the raw radio frequency signals acquired by energy absorption sensors operating in different frequency bands to obtain target filtered signals of different modes. Then, according to a preset error compensation model, the target filtered signals of different modes are corrected to obtain a target corrected signal. The target corrected signals of different modes are then weighted and summed to obtain a fused target signal. Based on the fused target signal, the average signal intensity attenuation rate in each sampling period is determined according to the number of samples taken in the horizontal and vertical polarization directions within each sampling period, resulting in an average signal intensity attenuation rate sequence. Finally, based on a preset conversion relationship, the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence can be converted into liquid sample moisture content to obtain a liquid sample moisture content sequence, thereby improving the accuracy of liquid sample moisture content detection to a certain extent.

[0065] Please see Figure 1 The present invention provides a method for detecting the water content of a liquid sample based on an energy absorption sensor. This method may include the following steps.

[0066] Step S110: Preprocess the raw radio frequency signals of different modes acquired by the energy absorption sensor to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point.

[0067] In this embodiment, "liquid sample" refers to a liquid sample, which may include liquid samples with good insulation and low conductivity, such as crude oil and finished oil.

[0068] In this embodiment, the energy absorption sensor is a device that detects based on the energy attenuation characteristics of radio frequency signals in liquid samples with different water contents. Different water contents of liquid samples result in different absorption capabilities of radio frequency signals. The sensor inversely deduces the water content by collecting the energy changes of the radio frequency signals.

[0069] In this embodiment, "modal" refers to the difference in the operating frequency band of the sensor (e.g., 100MHz, 200MHz, 300MHz). The raw radio frequency signal is the radio frequency energy attenuation signal directly acquired by the sensor without any processing. Different operating frequency bands correspond to different energy absorption sensors to achieve redundant deployment for measuring the water content of liquid samples. For example, energy absorption sensors operating in the 100MHz, 200MHz, and 300MHz frequency bands can be used to acquire the liquid signal intensity attenuation rate to obtain the raw radio frequency signals of these three frequency bands.

[0070] In this embodiment, each original radio frequency (RF) signal includes both horizontal and vertical polarization modes within each sampling period. Furthermore, for energy absorption sensors operating at different frequency bands, their polarization direction is the same at the same sampling point. In other words, for the same sampling time point, energy absorption sensors operating at 100MHz, 200MHz, and 300MHz all sample according to either horizontal or vertical polarization. Specifically, in this invention, the monopole antenna is 10mm in size. The transmitting circuit generates 100MHz, 200MHz, and 300MHz RF signals via a voltage-controlled oscillator, and these signals are coupled to the antenna feed point and transmitted via a low-noise amplifier. Because the antenna size is much smaller than the RF signal wavelength, the electromagnetic field generated by the RF signal operates in the near field and cannot propagate to the far field. At this time, the ambient medium in the near field (the monopole antenna is immersed in liquid) affects the electromagnetic field strength generated by the antenna in the near field. Higher water content results in a higher dielectric constant, leading to severe impedance mismatch in the near field and a reduction in the amplitude of the RF signal reflected on the antenna. For different transmission frequencies, the antenna impedance adaptation effect produced by the same liquid is also different.

[0071] In one specific embodiment, the energy absorption sensor adopts a dual-dimensional design of "3 frequency bands + 2 polarizations," equipped with a dual-polarization probe (switchable between horizontal and vertical polarization), and sets three differentiated operating frequency bands (100MHz, 200MHz, and 300MHz). Therefore, the energy absorption sensor can be configured to operate in the following modes: M1: 100MHz + horizontal polarization (suitable for high-viscosity liquid samples, such as crude oil, with a penetration depth ≥5mm); M2: 100MHz + vertical polarization (suitable for high-viscosity liquid samples containing suspended particles, enhancing particle moisture response); M3: 200MHz + horizontal polarization (suitable for medium-viscosity liquid samples, such as refined oil, balancing penetration and sensitivity); M4: 200MHz + vertical polarization (suitable for medium-viscosity emulsions, improving the accuracy of emulsion moisture identification); M5: 300MHz + horizontal polarization (suitable for low-viscosity liquid samples, such as chemical reagents, with fast signal response); M6: 300MHz + vertical polarization (suitable for low-viscosity liquid samples with high water content, offering the highest moisture response sensitivity). In this embodiment, preprocessing refers to a set of preliminary optimization operations on the original radio frequency signal, specifically including data cleaning and filtering, with the core objective of eliminating invalid interference information in the signal. Data cleaning may include missing value imputation, outlier handling, and environmental parameter normalization. The target filtered signal refers to the clean radio frequency signal after preprocessing, free from missing values, outliers, environmental interference, and noise, providing a reliable data foundation for subsequent error compensation and moisture content calculation.

[0072] Step S120: Based on the preset error compensation model, the target filtered signals of different modes are converted into target correction signals respectively; wherein, the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation.

[0073] In this embodiment, the preset error compensation model refers to the error correction mathematical model constructed in advance through experiments / calibration, which includes one or more of four types of compensation logic: flow rate-response compensation (correcting the timing deviation caused by flow rate interference), bubble interference compensation (correcting the signal interference caused by bubbles), sensor calibration error compensation (correcting the sensor's own hardware deviation), and emulsion layer interference compensation (correcting the additional absorption interference of the liquid sample emulsion layer on the radio frequency signal).

[0074] In this embodiment, the target correction signal refers to the target filter signal that, after error compensation, eliminates sensor hardware deviations and emulsion layer interference, thus further reducing system errors.

[0075] Step S130: Weighted summation of the target correction signals of different modalities to obtain the fused target signal; wherein the sum of the weights corresponding to the target correction signals of different modalities is 1.

[0076] In this embodiment, since the water content detection of a single sensor may have deviations under different modes, the accuracy of detecting the water content of liquid samples by using the same energy absorption sensor under different operating conditions is greatly improved compared to single-dimensional measurement.

[0077] In this embodiment, the fused target signal is the result obtained by weighted summation of the target correction signals of different modes. For example, the signal of the first mode is the first radio frequency signal collected by the energy absorption sensor in the 200MHz operating frequency band, and the signal of the second mode is the second radio frequency signal collected by the energy absorption sensor in the 300Hz operating frequency band. Based on the analysis of the first radio frequency signal, the second radio frequency signal, and the corresponding radio frequency signal under the standard oil sample, the weight of the first radio frequency signal is 0.6, and the weight of the second radio frequency signal is 0.4. Therefore, the fused target signal = 0.6 × first radio frequency signal + 0.4 × second radio frequency signal. It can be understood that in some other embodiments, the operating conditions of the energy absorption sensor in different modes can be greater than or equal to three, which can be set according to the specific oil sample.

[0078] In one specific embodiment, a 200MHz radio frequency signal is extended into a dual-mode switchable signal of 200MHz and 300MHz using frequency conversion technology. The energy attenuation coefficient slopes of the two signals are extracted, and a weighted fusion algorithm is used to assign weights to the real-time signal-to-noise ratio of the two modes. and And satisfy + =1, the fusion formula is as follows:

[0079]

[0080] The fused data improves the accuracy by 40% compared to single-mode measurement.

[0081] Understandably, in some other embodiments, the signal-to-noise ratio (SNR) and signal stability coefficient (CV, coefficient of variation) of the signal in each dimension can also be used. Here, SNR = effective signal amplitude / noise amplitude (noise amplitude is extracted from the signal frequency domain baseline). CV = signal standard deviation / signal mean (the smaller the CV, the more stable the signal). Then, the target corrected signal for each mode is dynamically adjusted. = ×( / Σ )×(1- ),in Let be the final weight of the i-th mode. Let be the real-time signal-to-noise ratio of the i-th mode. Let be the signal stability coefficient for the i-th mode. Finally, the target correction signals from different dimensions are fused. .in, This is the target correction signal for the i-th mode. It should be noted that when the calculated... In the case of this, the signal corresponding to that mode will not participate in the subsequent signal fusion calculation.

[0082] Step S140: Based on the fused target signal, determine the average signal strength attenuation rate in each sampling period according to the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, and obtain the average signal strength attenuation rate sequence.

[0083] In this embodiment, a sampling period can be determined based on the time taken for the horizontal and vertical polarization directions within that sampling period. Since the sampling time is taken at fixed time intervals, the number of samples can be used instead of the sampling time. Then, the weighting coefficient corresponding to the horizontal polarization direction can be determined according to the proportion of the number of samples taken in the horizontal polarization direction to the total number of samples within that sampling period. Similarly, the weighting coefficient corresponding to the vertical polarization direction can be determined according to the proportion of the number of samples taken in the vertical polarization direction to the total number of samples within that sampling period. Finally, the average RF signals in the horizontal and vertical polarization directions are weighted and summed based on their respective weighting coefficients to obtain the average signal strength attenuation rate for that sampling period.

[0084] In this embodiment, the average signal strength attenuation rate sequence is obtained by arranging the average signal strength attenuation rate of each sampling period according to the time series.

[0085] Step S150: Based on a preset conversion relationship, the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence is converted into liquid sample water content to obtain a liquid sample water content sequence.

[0086] In this embodiment, the liquid sample water content sequence is a set of water content data of each sampling point arranged according to the sampling period, which can reflect the real-time change trend of liquid sample water content (such as the dynamic fluctuation of crude oil water content in oil extraction).

[0087] In this embodiment, after obtaining the average signal intensity attenuation rate signal, it can be converted into the specific water content of the liquid sample according to the conversion law between the signal value and the liquid sample. The main parameters affecting the water content conversion include the dielectric constant of the liquid sample, the type of liquid sample, and the viscosity of the liquid sample. Based on a preset conversion model, the fused target signal can be converted into a liquid sample water content sequence.

[0088] In this embodiment, the preset conversion relationship can be achieved by first using a high-precision ADC to pre-acquire the radio frequency signal amplitudes of oil-water mixtures with different water contents from an energy absorption sensor. Subsequently, the relationship between water content and radio frequency signal amplitude is obtained using a lookup table method and linear interpolation, and the signal strength attenuation rate is expressed as the percentage decrease in amplitude. For example, data can be pre-acquired at a gradient of 1% water content from 1% to 99%, and finally, the value for 100% water can be acquired. Then, the water content range corresponding to the liquid sample can be determined first by using a lookup table method, and then the water content of the liquid sample can be determined within this range by linear interpolation.

[0089] Specifically, for example, for raw signals acquired from multiple frequency bands, first, missing signal values ​​are filled in and outliers are removed (such as jump values ​​caused by sudden interference). Then, the influence of environmental parameters such as temperature and pressure on the signal is eliminated (e.g., high temperature will enhance the absorption of radio frequency by the liquid sample, so it needs to be normalized to a standard environment). Then, random noise in the signal is filtered to finally obtain the target filtered signal. Then, a pre-calibrated model is called to perform error compensation to obtain the target corrected signal. Then, a precision weight is assigned to each modal signal, calculated according to "signal 1 × weight 1 + signal 2 × weight 2 + ..." to obtain the fused target signal. For the fused target signal, the average signal strength attenuation rate in each sampling period is calculated based on the proportion of samples in different polarization directions in each sampling period as weight coefficients. Finally, the water content in each sampling period is calculated through the "calibrated mapping relationship between average signal strength attenuation rate (radio frequency energy attenuation value) and water content" (e.g., the larger the attenuation value → the higher the water content), and the water content sequence is formed in chronological order.

[0090] In the above embodiments, multimodal signals can avoid the shortcomings of insufficient accuracy in detecting water content liquid samples using a single frequency band, thus improving detection coverage. Furthermore, preprocessing eliminates missing, abnormal, and environmental interference, providing clean data for subsequent calculations and reducing misjudgments of water content caused by signal noise. Secondly, the error compensation model specifically corrects for hardware deviations and the influence of the emulsion layer, reducing the interference of core error sources on the detection results.

[0091] In some embodiments, step S110 involves preprocessing the raw radio frequency signals of different modes acquired by the energy absorption sensor to obtain the target filtered signal, which may include the following steps.

[0092] Step S111: The original radio frequency signals of different modes are collected and cleaned to obtain standard radio frequency signals of each mode; wherein, the data cleaning includes filling missing values, handling outliers, and normalizing environmental parameters of the original radio frequency signals.

[0093] In this embodiment, data cleaning refers to the "set of sub-operations" of preprocessing, which optimizes the data consistency of the original radio frequency signal. This includes missing value filling (completing invalid data), outlier handling (removing interfering data), and environmental parameter normalization (eliminating the influence of environmental fluctuations). The core is to ensure that the original signals under different environments are on the same comparable benchmark.

[0094] In this embodiment, the standard radio frequency signal refers to a unified reference signal that has been cleaned to eliminate "data missing, outliers, and environmental differences" and can be directly used for subsequent filtering processing.

[0095] Step S112: Perform multi-layer filtering on the standard radio frequency signal to obtain the target filtered signal.

[0096] In this embodiment, multi-layer filtering refers to filtering different types of noise (such as burst impulse noise and high-frequency random noise) layer by layer through a combination of various filtering algorithms.

[0097] In one specific embodiment, for the original signal of each frequency band, the following steps are performed independently: missing value imputation → outlier handling → environmental parameter normalization. This involves first imputing missing data, then removing abrupt outliers, and finally converting the signal to values ​​under standard environmental conditions (e.g., 25°C, 0.1 MPa) to obtain standard RF signals for each mode (avoiding incomparability between different modes due to environmental differences). Then, a multi-level filtering algorithm is used to process the standard RF signals layer by layer (e.g., first filtering burst noise, then decomposing high-frequency noise, and finally smoothing the signal) to ultimately obtain the target filtered signal.

[0098] In the above embodiments, data cleaning unifies the benchmark for signals from different modalities and environments, avoiding signal deviations between modalities caused by environmental fluctuations (such as sudden temperature increases) or data loss, thus laying a comparable foundation for subsequent multimodal fusion. Multi-layer filtering specifically filters different types of noise (such as impulse noise and high-frequency noise). Compared with single filtering, the signal smoothness is higher, the noise residue is less, and the reliability of the target filtered signal is significantly improved.

[0099] In some embodiments, the signals acquired by the energy absorption sensor also include temperature signals and pressure signals. In step S111, the raw radio frequency signals of different modes acquired are cleaned to obtain standard radio frequency signals of each mode, which may include the following steps.

[0100] Step S1111: If the number of consecutive missing data in the original radio frequency signal is less than or equal to the preset number of samplings, the missing region is linearly interpolated using the values ​​adjacent to the missing region to obtain the initial repair signal.

[0101] In this embodiment, identifying missing data segments in the original signal (such as signal blanks caused by a brief disconnection of the sensor) can be achieved by performing time continuity detection on the original oil signal (such as the 200MHz radio frequency energy attenuation value) collected by the sensor, identifying missing values ​​caused by communication interruption, and marking them as "NaN". Then, these missing values ​​are filled in using a preset method to obtain a first radio frequency attenuation signal without missing data.

[0102] In this embodiment, the first radio frequency attenuation signal refers to the radio frequency signal that has only been filled with missing values, which completes the missing data, but there may still be outliers and environmental interference.

[0103] In this embodiment, the preset number of samplings is a pre-set threshold for switching missing value filling methods (such as 3 samplings) to distinguish between "short-term missing values" and "long-term missing values" and avoid the limitations of a single filling method.

[0104] In this embodiment, linear interpolation is used for a simple filling method for short-term missing data—utilizing the effective values ​​before and after the missing region to calculate the missing point value according to a linear relationship (e.g., if the value before the missing point is 10dB, the value after is 12dB, and the missing value in the middle is filled with 11dB). For example, for short-term missing data (≤3 sampling points), the "adjacent time signal interpolation method" is used, based on the continuity of the energy absorption signal (e.g., slow changes at the oil-water interface), using the linear interpolation formula:

[0105]

[0106] In the formula, Indicates the missing values ​​in the missing region. This represents the value of the previous sampling point in the currently missing region. This represents the value of the next sampling point in the currently missing region. This indicates the position of the missing value within the current missing region.

[0107] In this embodiment, the initial repair signal refers to the temporary signal obtained by linear interpolation of the long-term missing area, which serves as the basis for subsequent accurate filling (to avoid excessive error from direct prediction).

[0108] Step S1112: When the number of consecutive missing data in the original radio frequency signal is greater than the preset number of samplings, the initial repair signal is predicted and filled using a pre-trained long short-term memory network model based on the changing trend of historical data corresponding to the missing region and the initial repair signal to obtain the first radio frequency attenuation signal.

[0109] In this embodiment, the pre-trained Long Short-Term Memory (LSTM) network model refers to an artificial intelligence model that can learn the trends of time-series data—trained in advance using historical radio frequency signal data, it can predict the signal change trend in long-term missing regions (e.g., when the water content of a liquid sample slowly increases, the signal attenuation value should gradually increase). Specifically, when more than three consecutive missing data samples are found, the LSTM predicts and fills in these missing values.

[0110] It is understood that, in some other embodiments, the long-term missing data can also be filled by deploying a backup sensor in the operating frequency band corresponding to the missing value and using the data from the backup sensor.

[0111] In this embodiment, the first radio frequency attenuation signal refers to the signal without missing values ​​after being filled with missing values ​​in different scenarios (short-term linear interpolation, long-term LSTM prediction), and the filling accuracy is higher than that of a single method.

[0112] In the above embodiments, linear interpolation (efficient and sufficiently accurate) is used for short-term missing data, and LSTM (using historical trends to improve prediction accuracy) is used for long-term missing data. This avoids wasting computing power by using complex models for short-term missing data and causing bias by using simple interpolation for long-term missing data, thus balancing filling efficiency and accuracy.

[0113] S1113: Calculate the average value and standard deviation of the first radio frequency attenuation signal.

[0114] In this embodiment, the average value and standard deviation are statistical characteristic values ​​of the first radio frequency attenuation signal—the average value reflects the overall level of the signal, and the standard deviation reflects the degree of signal fluctuation. The two are combined to determine abnormal values ​​that deviate from the normal range.

[0115] S1114: The absolute value of the difference between the value in the first radio frequency attenuation signal and the average value is greater than the preset standard deviation, or the value in the first radio frequency attenuation signal is greater than the limit value of liquid sample energy absorption, is determined as an outlier; wherein, the preset standard deviation is the product of the standard deviation and the first preset multiple.

[0116] In this embodiment, the preset standard deviation refers to a pre-set threshold for judging outliers, which is "standard deviation × first preset multiple" (e.g., 3 times the standard deviation). Values ​​exceeding this threshold are considered outliers that significantly deviate from normal fluctuations. The limit value of liquid sample energy absorption is the upper limit of the maximum absorption capacity of a liquid sample (e.g., crude oil, chemical solution) for radio frequency signals (e.g., the maximum attenuation value of a certain crude oil is 20dB). Signals exceeding this value are necessarily interference (e.g., signal saturation caused by equipment failure).

[0117] S1115: Take the abnormal point corresponding to the abnormal value as the center of the first preset window, and determine the median of the values ​​in the first preset window as the value of the abnormal point to obtain the second radio frequency attenuation signal.

[0118] In this embodiment, the first preset window refers to a local data interval (such as 5 sampling points) centered on the outlier, used to extract local normal signal features (median) and avoid the influence of a single outlier. The median refers to the middle value after the values ​​in the window are sorted by size, and it is more "resistant to extreme values" than the average value (if there is one jump value in the window, the median can still reflect the normal level).

[0119] In this embodiment, the second radio frequency attenuation signal refers to the radio frequency signal that has undergone missing value filling and outlier processing, which eliminates missing and abnormal interference, but is still affected by environmental parameters.

[0120] In a specific embodiment, for sudden signal jumps caused by sensor vibration or scaling (such as a 50% sudden increase in absorption value at a certain moment), a "dual threshold detection method" is used: For the first RF attenuation signal after filling, the average value (μ) and standard deviation (σ) of all valid sampling points are statistically analyzed to determine the normal fluctuation range (such as μ ± 3σ). If a value satisfies "|value - μ| > preset standard deviation (such as 3σ)" or "value > liquid sample energy absorption limit value", it is determined to be an abnormal value (the former is a fluctuation abnormality, and the latter is a physical limit abnormality). Then, a first preset window (such as 2 sampling points before and after, for a total of 5 points) is defined with the abnormal point as the center, and the median of the values ​​within the window is calculated. The abnormal point value is replaced with this median to obtain the second RF attenuation signal without abnormalities. Thus, two types of abnormal values ​​can be accurately identified, covering both interference exceeding normal fluctuations (such as electromagnetic pulses) and fault signals exceeding physical limits (such as sensor overload), avoiding the omission of abnormal values. In addition, replacing with the window's median value can effectively eliminate other potential minor interferences within the window, ensuring that the replaced value matches the local normal signal level and further improving signal reliability.

[0121] Step S1116: Based on the temperature signal and the pressure signal, convert the energy attenuation values ​​of each sampling point in the second radio frequency attenuation signal into target energy attenuation values ​​under standard conditions to obtain a standard radio frequency signal.

[0122] In this embodiment, based on the calibration relationship between temperature / pressure and radio frequency attenuation (e.g., for every 10°C increase in temperature, the attenuation value increases by 0.5dB), the value of each sampling point of the second radio frequency attenuation signal is corrected to the corresponding value under standard conditions to obtain a standard radio frequency signal, thereby eliminating the interference of temperature and pressure on the energy absorption signal. The original signal is normalized to the standard environment (25°C, 0.1MPa) using the following formula:

[0123]

[0124] in, For standard radio frequency signals, This is the second radio frequency attenuation signal. , These are temperature and pressure correction factors calibrated through experiments (based on a pre-set coefficient library for different oil products).

[0125] In the above embodiments, missing value imputation avoids interruptions in moisture content calculation due to data gaps, ensuring detection continuity (e.g., no data gaps in online oil pipeline monitoring). The median replaces outliers (more resistant to extreme values ​​than the mean), avoiding misjudgments of moisture content jumps caused by single outliers (e.g., mistakenly identifying interference values ​​as a sudden increase in moisture content). Temperature / pressure normalization offsets the impact of environmental changes on the signal, ensuring comparability of detection results under different environments (e.g., consistent signal baseline between high summer temperatures and low winter temperatures).

[0126] In some embodiments, step S112, performing multi-layer filtering on the standard radio frequency signal to obtain the target filtered signal, may include the following steps.

[0127] Step S1121: In the standard radio frequency signal, the center value of the second preset window is replaced with the median of the second preset window to obtain the first filtered signal; wherein, the size of the second preset window when the liquid sample is in a stable period is larger than the size of the second preset window when the liquid sample is in a sudden change period.

[0128] In this embodiment, the second preset window refers to a local data window used for median filtering. The window size (number of sampling points) is dynamically adjusted according to the liquid sample state—a large window is used when the liquid sample is stable (slow change in water content) and a small window is used when the liquid sample undergoes a sudden change (rapid change in water content). Specifically, the stable liquid sample period refers to the period when the liquid sample water content changes slowly (such as when a chemical storage tank is stationary), and the signal fluctuation is small, requiring a large window filtering (such as 10 sampling points) to enhance the smoothing effect; the sudden liquid sample period refers to the period when the liquid sample water content changes abruptly (such as when switching oil wells in an oil pipeline), and the signal changes rapidly, requiring a small window filtering (such as 3 sampling points) to avoid over-filtering and masking the true sudden change.

[0129] In this embodiment, the first filtered signal refers to the signal that has undergone dynamic median filtering, which mainly suppresses signals containing sudden jumps (such as spikes caused by vibration and impact).

[0130] Step S1122: Perform wavelet packet decomposition on the first filtered signal to obtain the second filtered signal.

[0131] In this embodiment, the wavelet packet decomposition-multi-scale signal decomposition algorithm can decompose the first filtered signal into "high-frequency components" (random noise) and "low-frequency components" (effective signal). By retaining the low-frequency components and removing the high-frequency components, the second filtered signal is obtained, mainly eliminating "high-frequency random noise". Specifically, for periodic interference at the 50Hz power frequency, wavelet packet decomposition filtering is used to eliminate it. The signal is decomposed into 5 layers of wavelet packets (using the dB4 wavelet), the frequency band corresponding to the noise is located in the frequency domain and set to zero, and after reconstructing the signal, the periodic noise can be attenuated by ≥40dB.

[0132] Step S1123: Smooth the second filtered signal using the Kalman filter algorithm to obtain the target filtered signal.

[0133] In this embodiment, the Kalman filter algorithm is a dynamic signal smoothing algorithm that can dynamically adjust the weights based on the "predicted value" and "measured value" of the signal, smoothing signal fluctuations while preserving the true trend of change, and finally obtaining the target filtered signal. Specifically, for random noise (such as residual Gaussian white noise), "adaptive Kalman filtering" is used for dynamic smoothing: (1) Establishing the state equation: regard the change in water content as a linear dynamic system. ,in (2) Observation equation: based on the mapping relationship between energy absorption value and water content. ,in To observe noise; (3) Update the noise covariance in real time ( , This allows the filter gain to adapt to signal fluctuations, achieving a noise suppression rate of up to 92% when the signal-to-noise ratio is ≥10dB.

[0134] In the above embodiments, the dynamic window mid-range filtering balances the smoothing effect during stable conditions with signal fidelity during abrupt changes, avoiding the problems of large windows masking abrupt changes and small windows failing to provide sufficient smoothing. Wavelet packet decomposition specifically removes high-frequency noise, and Kalman filtering further smooths small fluctuations. After two layers of filtering, there is very little residual noise in the signal, providing a high-purity data foundation for subsequent error compensation.

[0135] In some embodiments, step S120, converting target filtered signals of different modes into target correction signals according to a preset error compensation model, may include the following steps.

[0136] Step S121: Real-time acquisition of liquid sample flow rate data; calculation of signal delay duration corresponding to the liquid sample flow rate data according to a preset mapping relationship model; and shifting the target filtered signal at the current moment forward based on the signal delay duration to obtain a time-series target correction signal; wherein, the sampling frequency of the liquid sample flow rate data is the same as the sampling frequency of the original radio frequency signal.

[0137] In this embodiment, a miniature flow meter is installed at the inlet of the detection chamber to collect the liquid sample flow rate v (unit: m / s) in real time, and the sampling frequency is consistent with the radio frequency signal (1kHz).

[0138] In this embodiment, 3-5 typical flow rates (e.g., 0.1 m / s, 0.5 m / s, 1.0 m / s, 2.0 m / s, 3.0 m / s) are selected. At each flow rate, a standard liquid sample with a known water content is injected into the detection chamber, and the delay Δt (unit: ms) between the radio frequency signal and the actual liquid sample reaching the probe is recorded. With flow rate v as the independent variable and delay Δt as the dependent variable, a mapping relationship is obtained through linear fitting: Δt = a × v + b, where a and b are calibration coefficients (e.g., a = 5 ms·s / m, b = 0.2 ms, specifically determined experimentally). A "flow rate-delay" mapping table is formed and stored.

[0139] In this embodiment, based on the real-time collected flow velocity v, the signal delay Δt corresponding to the current flow velocity is obtained through linear interpolation using a mapping table. The radio frequency signal at the current moment is shifted forward by Δt timestamps to synchronize with the actual arrival time of the liquid sample at the probe, thus completing the timing deviation correction.

[0140] Step S122: Perform frequency domain transformation on the time-series target correction signal to identify the bubble signal amplitude; perform wavelet decomposition on the time-series target correction signal, and determine the low-frequency component representing the effective signal and the high-frequency component representing the bubble interference based on the bubble signal amplitude; set the portion of the high-frequency component whose amplitude exceeds the preset amplitude to 0, and obtain the bubble target correction signal through inverse wavelet transformation.

[0141] In this embodiment, a Fast Fourier Transform (FFT) is performed on the time-series target correction signal after flow velocity compensation to obtain the signal's frequency domain spectrum. The interference of bubbles on the radio frequency signal manifests as discrete spikes in the frequency domain (discontinuous distribution, spike amplitude ≥ 20% of the signal's fundamental amplitude). A threshold method is used to identify the frequency points corresponding to these spikes. If more than three frequency points exist within a single period, bubble interference is considered to exist, and bubble interference compensation is required.

[0142] In this embodiment, the db4 wavelet is selected as the base wavelet, and the time-series target correction signal is decomposed into three levels of wavelet decomposition to obtain low-frequency components (effective signals) and high-frequency components (including bubble interference). For the high-frequency components containing bubble interference, an adaptive preset amplitude is set. ,in, Let N be the noise standard deviation of the high-frequency components, and N be the signal length. Then, for high-frequency components with amplitudes exceeding [a certain value], [the following is a list of parameters]. The part (bubble interference) is set to zero, the effective high-frequency components below λ are retained, and the signal is reconstructed by wavelet inverse transform to obtain the pure signal after removing bubble interference.

[0143] Step S123: Based on the thickness of the emulsion layer in the liquid sample and the emulsion layer thickness correction factor, the bubble target correction signal is corrected to obtain the emulsion layer interference correction signal.

[0144] In this embodiment, the emulsion layer refers to an unstable layer formed by the mixture of oil, water, and impurities in the liquid sample (such as an oil-water emulsion layer in crude oil extraction), which will additionally absorb radio frequency signals, resulting in an overestimation of the measured value. The emulsion layer thickness correction factor is a pre-calibrated "thickness-interference coefficient" (e.g., for every 1 mm increase in thickness, the attenuation value increases by 0.1 dB), used to quantify the interference level of the emulsion layer.

[0145] In this embodiment, after the bubble interference compensation is performed, an "emulsion layer thickness correction factor" is introduced to address the absorption signal distortion caused by the emulsion layer. The correction formula is as follows:

[0146]

[0147] Where d is the thickness of the emulsion layer, and the correction factor of 0.02 is determined experimentally.

[0148] Understandably, in some other embodiments, the emulsion thickness correction factor is also related to the flow rate. The higher the flow rate, the lower the emulsion thickness correction factor can be by 10%-20%, which can be determined experimentally.

[0149] Step S124: Based on the preset system error model, the emulsion layer interference correction signal is converted into a target correction signal; the preset system error compensation model is used to calibrate the error of the energy absorption sensor; the preset system error compensation model is a linear transformation model constructed by taking the radio frequency signal of the standard liquid sample as the dependent variable and the radio frequency signal measured by the energy sensor on the standard liquid sample as the independent variable.

[0150] In this embodiment, the preset system error model refers to the calibration model for hardware deviations of the energy absorption sensor. After the sensor leaves the factory, there may be hardware drift (such as radio frequency transmission power deviation), which needs to be corrected by standard liquid sample calibration. The standard liquid sample refers to the "reference liquid sample" with known accurate water content (such as standard crude oil with water content of 10%, 30%, and 50%), and its radio frequency signal is the "true reference value".

[0151] In this embodiment, the linear transformation model is a common form of the systematic error model, with an expression such as "y = ax + b" (y is the actual radio frequency signal of the standard liquid sample, x is the radio frequency signal measured by the sensor, and a and b are correction coefficients determined by least squares fitting based on the calibration data), which is used to convert the measured value into the true value.

[0152] In the above embodiments, flow velocity interference causes signal timing deviations, which need to be corrected first to ensure the time synchronization of subsequent signals; bubble interference is a type of signal superposition interference and needs to be stripped before emulsion layer compensation (to avoid mutual interference between the two types of superposition interference); sensor calibration error is an inherent hardware deviation and is corrected last to ensure accuracy. Therefore, by determining the compensation order based on the characteristics of the interference sources (timing deviation type → signal superposition type → system hardware type): flow velocity-response hysteresis compensation → bubble interference compensation → emulsion layer interference compensation → sensor calibration error compensation, the accuracy of the target corrected signal is significantly higher than that of the uncompensated signal, providing a high-precision signal source for subsequent multimodal fusion.

[0153] It is understandable that in some other embodiments, compensation for certain types of errors may not necessarily be necessary. Therefore, error compensation can be performed simply by following the processing logic sequence of flow rate-response hysteresis compensation → bubble interference compensation → emulsion layer interference compensation → sensor calibration error compensation. For example, in the absence of bubble interference compensation, emulsion layer interference compensation can be directly performed on the hysteresis result of flow rate-response hysteresis compensation.

[0154] In some embodiments, after step S130, the liquid sample water content detection method based on the energy absorption sensor may further include the following steps.

[0155] Step S1311: The fusion target signal at the current moment and the fusion target signal at the previous moment are weighted and summed using a second preset weighting coefficient to obtain a target stable signal; wherein, the sum of the weights of the fusion target signal at the current moment and the weights of the fusion target signal at the previous moment is 1; wherein, when the liquid sample is in a stable state, the weighting coefficient corresponding to the fusion target signal at the current moment is less than the weighting coefficient corresponding to the fusion target signal at the current moment when the liquid sample is in an unstable state.

[0156] In this embodiment, the second preset weight coefficient refers to the weight used for time-series signal fusion (current time target signal weight + previous time weight = 1), which is dynamically adjusted according to the liquid sample state—when the liquid sample is stable, the previous time is emphasized (to avoid small fluctuations), and when the liquid sample is unstable, the current time is emphasized (to track real changes).

[0157] In this embodiment, the target stable signal refers to the signal that has undergone multimodal fusion and temporal fusion, combining the full-range advantages of multimodal fusion with the stability of temporal fusion, thus avoiding random fluctuations in the signal at a single moment. Specifically, for example, dynamic exponential smoothing uses exponential moving average (EMA):

[0158]

[0159] in, The smoothing factor is dynamically adjusted by monitoring the rate of change of water content ΔW between adjacent time points. For example, taking in the stationary section =0.3 enhances smoothness, taking at the mutation segment =0.8 preserves response speed.

[0160] Accordingly, in step S140, based on the fused target signal, the average signal strength attenuation rate in the current sampling period is determined according to the number of samples in the horizontal polarization direction and the vertical polarization direction in the current sampling period, which may include the following steps.

[0161] Step S1401: Based on the target stable signal, determine the average signal strength attenuation rate in the current sampling period according to the number of samples in the horizontal polarization direction and the vertical polarization direction in the current sampling period, and obtain the average signal strength attenuation rate sequence.

[0162] In this embodiment, the original fused target signal is replaced with the target stable signal, and the average signal strength attenuation rate of each sampling period is calculated to obtain a more stable average signal strength attenuation rate sequence.

[0163] In the above embodiments, time-series fusion smooths current fluctuations using signals from the previous moment, avoiding abrupt changes in the water content sequence caused by random noise at a single moment. Furthermore, it prioritizes smoothing (ensuring accuracy) during stable periods and prioritizes the current signal (avoiding lag) during unstable periods, balancing stability and response speed. Secondly, the target stable signal possesses both the advantages of multimodal full-range representation and temporal stability, resulting in a water content sequence that more closely reflects the true state of the liquid sample.

[0164] In some embodiments, in step S140, based on the fused target signal, the average signal strength attenuation rate in each sampling period is determined according to the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, and an average signal strength attenuation rate sequence is obtained, which may include the following steps.

[0165] Step S141: For each sampling period, calculate the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction in the current sampling period based on the target fused signal; or, calculate the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction in the current sampling period based on the target stable signal.

[0166] In this embodiment, the average level signal strength attenuation rate refers to the average value P obtained by horizontal polarization within the sampling period. H The average vertical signal strength attenuation rate refers to the average value P obtained through vertical polarization within the sampling period. V .

[0167] Step S142: Determine the horizontal polarizability and vertical polarizability based on the number of samples taken in the horizontal polarization direction and the number of samples taken in the vertical polarization direction within the current sampling period.

[0168] In this embodiment, horizontal polarizability refers to the ratio of the number of sampling points with horizontal polarization to the total number of samples within the sampling period. Vertical polarizability refers to the ratio of the number of sampling points with vertical polarization to the total number of samples within the sampling period. Therefore, horizontal polarizability can be expressed as: Vertical polarizability can be expressed as: .

[0169] Understandably, in a typical sampling period, only one batch of horizontal polarization and one batch of vertical polarization are performed. Therefore, the horizontal polarizability can be calculated by using the proportion of horizontal polarization time to the sampling period, and the vertical polarizability can be calculated by using the proportion of vertical polarization time to the sampling period.

[0170] Step S143: Add the product of the horizontal polarization and the average level signal strength attenuation rate, and the product of the vertical polarization and the average vertical signal strength attenuation rate, to obtain the average signal strength attenuation rate of the current sampling period.

[0171] In this embodiment, the average signal strength attenuation rate reflects the average change in water content during the sampling period. The average signal strength attenuation rate can be expressed as: .

[0172] Step S144: Concatenate the average signal strength attenuation rates of each sampling period to obtain an average signal strength attenuation rate sequence.

[0173] In this embodiment, the average signal strength attenuation rate of each sampling period is spliced ​​together in chronological order to form an average signal strength attenuation rate sequence.

[0174] In some embodiments, after step S150, the liquid sample water content detection method based on the energy absorption sensor may further include the following steps.

[0175] Step S161: Calculate the sub-mean and sub-standard deviation of the liquid sample water content in each sampling period within the third preset window in the liquid sample water content sequence.

[0176] In this embodiment, the third preset window is used to analyze the time window of local anomalies in the moisture content sequence (such as 10 sampling points), and the local anomalies are determined by the statistical features (sub-mean, sub-standard deviation) within the window.

[0177] In this embodiment, the sub-mean / sub-standard deviation are local statistical values ​​of moisture content data within the third preset window—the sub-mean reflects the local moisture content level, and the sub-standard deviation reflects the degree of local fluctuation. The preset sub-standard deviation is the threshold for judging local anomalies, which is "sub-standard deviation × second preset multiple" (e.g., 2 times the sub-standard deviation). Moisture content exceeding this threshold is considered a local anomaly.

[0178] Step S162: If the absolute value of the difference between the liquid sample water content and the sub-mean value in the water content sequence for m consecutive sampling periods is greater than the preset sub-standard deviation, the water content of the current sampling period is replaced with the average value of the water content of the previous p sampling periods of the current sampling point; wherein, the preset sub-standard deviation is the product of the sub-standard deviation and the second preset multiple.

[0179] Step S163: If the absolute value of the difference between the liquid sample water content and the sub-mean value at n consecutive sampling points in the water content sequence is greater than the preset sub-standard deviation, the health diagnosis of the energy absorption sensor is triggered; where n > m.

[0180] In this embodiment, the ratio of m consecutive sampling points to n consecutive sampling points serves as the threshold for classifying the degree of anomaly (n > m, e.g., m = 2, n = 5). m corresponds to minor, short-term anomalies (which can be corrected), while n corresponds to severe, long-term anomalies (requiring investigation of sensor malfunction). Sensor health diagnosis refers to the fault diagnosis mechanism for the energy absorption sensor (e.g., checking for sensor misalignment, hardware malfunction, or signal line interference). This mechanism is triggered when a severe anomaly occurs, preventing continuous misjudgments caused by sensor malfunction.

[0181] In one specific embodiment, a "sliding window confidence interval" (window size = 10 sampling periods) is established, and the mean μ and standard deviation σ of the moisture content within the window are calculated in real time. When the current value exceeds [ -2 , +2 When the value exceeds the limit, record the number of times. If it exceeds the limit twice consecutively, it is judged as a "potential anomaly" and a "smooth value with warning" (replaced with the average of the previous 3 values) is output; if it exceeds the limit 5 times consecutively, "sensor health diagnosis" is triggered (checking for scaling or malfunction).

[0182] In the above embodiments, replacing m consecutive abnormal points with an average value avoids production errors caused by short-term interference (such as momentary electromagnetic interference) (e.g., misjudging moisture content and causing shutdown). Furthermore, in the case of n consecutive abnormal points, a health check is triggered to promptly detect sensor hardware faults (such as offset or damage) or circuit interference, preventing faulty sensors from continuously outputting erroneous data and ensuring the long-term reliability of the detection system. In other words, minor anomalies are self-corrected (improving detection efficiency), while serious anomalies are manually investigated (ensuring system safety), balancing automation and safety to meet the actual needs of industrial scenarios.

[0183] It is understood that in the above embodiments, parameters can be adaptively adjusted according to operating conditions to optimize algorithm parameters based on the characteristics of radio frequency energy absorption (sensitive to dielectric constant, attenuation positively correlated with water content) for different moisture content ranges and emulsion layer states. Furthermore, parameters can be automatically switched by real-time monitoring of features such as moisture content gradient and signal-to-noise ratio (SNR), triggering a parameter switching mechanism. Simultaneously, the emulsion layer can be identified through the "attenuation-phase joint characteristics" of the radio frequency signal.

[0184] In some embodiments, the energy absorption sensor is an array sensor probe, which includes multiple sub-energy absorption sensors, including normal region probes and abnormal region probes; each sub-energy absorption sensor is configured to acquire raw radio frequency signals of different modes. After step S110, the liquid sample water content detection method based on the energy absorption sensor may further include the following steps.

[0185] Step S1201: For each mode of the target filtered signal in each of the sub-energy absorption sensors, calculate the deviation rate between the target filtered signal and the array average signal corresponding to the target filtered signal.

[0186] In this embodiment, array-type sensor probes are distributed and installed within the detection chamber, covering the entire "center + perimeter" area of ​​the liquid sample flow / storage region. Specifically, nine sub-energy absorption sensors are selected, comprising a total of 27 raw radio frequency signals, which are processed according to the method described in the above embodiment to obtain their target filtered signals.

[0187] In this embodiment, for each dimension, the deviation rate .in, Let j be the signal value of the i-th probe in the j-th dimension. Let be the average signal value of the j-th dimension of each sub-energy absorption sensor.

[0188] Step S1202: If the deviation rate of k consecutive sampling points is greater than or equal to a preset deviation, the sub-energy absorption sensor corresponding to the target filtered signal is determined to be an abnormal region probe.

[0189] In this embodiment, a preset deviation is set. The default value is 5%, which can be adjusted according to the homogeneity of the liquid sample, such as increasing it to 8% under turbulent conditions. If the deviation rate of a certain sub-energy absorption sensor... If the deviation rate of k consecutive sampling points remains in this state, then the area corresponding to the sub-energy absorption sensor is determined to be a local abnormal area, and the sub-energy absorption sensor is an abnormal area probe. The remaining sub-energy absorption sensors are normal area probes.

[0190] Step S1203: Assign different weight coefficients to the normal region probe and the abnormal region probe respectively, and perform spatial fusion in the same dimension to obtain the target fusion filtered signal.

[0191] In this embodiment, the spatial weight of the normal region probe can be set to 1. For the abnormal region probe, the spatial weight can be reduced, such as... This avoids anomalous signals dominating the fusion result while retaining some reference value. The spatial fusion signal of the j-th mode. This allows us to obtain the target fused filter signal for that mode. Let be the spatial weight of the i-th sub-energy absorption sensor.

[0192] In this embodiment, after spatial fusion, the multiple signals from each sub-energy absorption sensor are finally integrated into one signal corresponding to each mode, which is used for subsequent error compensation processes.

[0193] Please see Figure 2 In other embodiments, during the measurement of the liquid sample by the energy absorption sensor, the original sensor signal can be processed as described above to output the specific result of the water content in the oil. However, this method can also evaluate the algorithm effect through quantitative indicators, continuously iterate and optimize parameters, realize algorithm iterative updates, and further improve detection accuracy and reduce error.

[0194] In this embodiment, the performance of the water content sequence generated by any of the above-mentioned energy absorption sensor-based liquid sample water content detection methods can be evaluated using three dimensions: accuracy, stability, and dynamic response. The accuracy index can be determined by comparing the root mean square error (RMSE) and maximum absolute error (MAE) with the measured values ​​of a standard oil sample. The stability index can be judged by the standard deviation (Std) and fluctuation range (Max-Min) of the measurement results within one hour. The dynamic response index can be determined by the response time during abrupt changes at the oil-water interface (time from signal change to output stabilization < 500ms).

[0195] Then, if all three dimensions meet the standards, it indicates that the current liquid sample moisture content detection model based on the energy absorption sensor conforms to the actual results and requires no adjustment. If the standards are not met, i.e., when the RMSE is >0.8% for three consecutive times under a certain operating condition, "parameter optimization" is triggered, using a particle swarm optimization algorithm to optimize parameters such as the filtering window and fusion weights. Alternatively, after accumulating 100 sets of operating condition data, a "operating condition-parameter" mapping model is automatically trained to achieve intelligent pre-adjustment of parameters.

[0196] In addition, the calibration curve coefficients can be automatically updated every 24 hours with newly collected calibration data (standard oil samples are injected periodically).

[0197] With dynamic parameter adjustment, it can adapt to the full range of moisture content measurement from 0 to 100%, and operate stably in environments of -40℃ to 150℃ and 0-150MPa.

[0198] Please see Figure 3 One embodiment of the present invention provides a liquid sample moisture content detection device based on an energy absorption sensor. The liquid sample moisture content detection device based on an energy absorption sensor may include: a signal preprocessing module, an error correction module, a signal fusion module, a signal intensity attenuation rate calculation module, and a moisture content conversion module.

[0199] The signal preprocessing module is used to preprocess the raw radio frequency signals of different modes acquired by the energy absorption sensor to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point.

[0200] The error correction module is used to convert target filtered signals of different modes into target corrected signals according to a preset error compensation model; wherein the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation.

[0201] The signal fusion module is used to perform weighted summation of target correction signals of different modalities to obtain a fused target signal; wherein the sum of the weights corresponding to the target correction signals of different dimensions is 1.

[0202] The signal strength attenuation rate calculation module is used to determine the average signal strength attenuation rate in each sampling period based on the fused target signal and the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, so as to obtain the average signal strength attenuation rate sequence.

[0203] The moisture content conversion module is used to convert the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence into the liquid sample moisture content based on a preset conversion relationship, so as to obtain the liquid sample moisture content sequence.

[0204] The specific functions and effects of the liquid sample moisture content detection device based on the energy absorption sensor can be explained by referring to other embodiments in this specification, and will not be repeated here. Each module in the liquid sample moisture content detection device based on the energy absorption sensor can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0205] Please see Figure 4 One embodiment of the present invention can provide an electronic device, the electronic device comprising:

[0206] A memory, and one or more processors communicatively connected to the memory;

[0207] The memory stores instructions that can be executed by the one or more processors, which, when executed by the one or more processors, enable the one or more processors to implement the liquid sample moisture content detection method based on the energy absorption sensor described in any of the above embodiments.

[0208] One embodiment of the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the liquid sample moisture content detection method based on an energy absorption sensor as described in any of the above embodiments.

[0209] The embodiments of this specification also provide a computer program product containing instructions that, when executed by a computer, cause the computer to perform the liquid sample water content detection method based on an energy absorption sensor as described in any of the above embodiments.

[0210] It is understood that the specific examples in this document are only intended to help those skilled in the art better understand the embodiments described herein, and are not intended to limit the scope of the invention.

[0211] It is understood that in the various embodiments described in this specification, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments described in this specification.

[0212] It is understood that the various implementation methods described in this specification can be implemented individually or in combination, and the implementation methods in this specification are not limited in this respect.

[0213] Unless otherwise stated, all technical and scientific terms used in the embodiments of this specification have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this specification. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items. The singular forms "a," "the," and "the" as used in the embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0214] It is understood that the processor in this invention can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method implementation can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this specification. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this specification can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0215] It is understood that the memory in this invention can be volatile memory or non-volatile memory, or may include both. Specifically, the non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM). It should be noted that the memory in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0216] The user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in the embodiments of the present invention are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0217] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0218] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the aforementioned method implementations, and will not be repeated here.

[0219] In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0220] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0221] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0222] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of this specification, in essence, or the parts that contribute to the prior art, or parts of the technical solutions, can be embodied in the form of software products. These computer software products are stored in a storage medium and include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0223] The above description is merely a specific embodiment of this specification, but the scope of protection of this invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this specification should be included within the scope of protection of this specification. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A method for detecting the water content of a liquid sample based on an energy absorption sensor, characterized in that, The method includes: The raw radio frequency signals of different modes acquired by the energy absorption sensor are preprocessed to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point; According to the preset error compensation model, the target filtered signals of different modes are converted into target correction signals respectively; wherein, the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation. The target correction signals of different modalities are weighted and summed to obtain the fused target signal; wherein the sum of the weights corresponding to the target correction signals of different modalities is 1. Based on the fused target signal, the average signal strength attenuation rate in each sampling period is determined according to the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, thus obtaining the average signal strength attenuation rate sequence. Based on a preset conversion relationship, the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence is converted into liquid sample water content to obtain a liquid sample water content sequence.

2. The method according to claim 1, characterized in that, The raw radio frequency signals of different modes acquired by the energy absorption sensor are preprocessed to obtain the target filtered signal, including: The raw radio frequency signals of different modes collected are cleaned to obtain standard radio frequency signals for each mode; wherein, the data cleaning includes filling missing values, handling outliers, and normalizing environmental parameters of the raw radio frequency signals. The standard radio frequency signal is subjected to multi-layer filtering to obtain the target filtered signal.

3. The method according to claim 2, characterized in that, The energy absorption sensor also collects temperature and pressure signals. The raw radio frequency signals of different modes are cleaned to obtain standard radio frequency signals for each mode, including: If the number of consecutive missing data in the original radio frequency signal is less than or equal to the preset number of samplings, the missing region is linearly interpolated using the value adjacent to the missing region to obtain the initial repair signal. If the number of consecutive missing data in the original radio frequency signal is greater than the preset number of samplings, the initial repair signal is predicted and filled using a pre-trained long short-term memory network model based on the changing trend of historical data corresponding to the missing region and the initial repair signal to obtain the first radio frequency attenuation signal. Calculate the average value and standard deviation of the first radio frequency attenuation signal; An outlier is defined as a value where the absolute value of the difference between the value in the first radio frequency attenuation signal and the average value is greater than a preset standard deviation, or a value in the first radio frequency attenuation signal is greater than the limit value of liquid sample energy absorption; wherein, the preset standard deviation is the product of the standard deviation and a first preset multiple. The abnormal point corresponding to the abnormal value is taken as the center of the first preset window, and the median of the values ​​in the first preset window is determined as the value of the abnormal point to obtain the second radio frequency attenuation signal. Based on the temperature signal and the pressure signal, the energy attenuation values ​​of each sampling point in the second radio frequency attenuation signal are converted into target energy attenuation values ​​under standard conditions to obtain a standard radio frequency signal.

4. The method according to claim 2, characterized in that, The standard radio frequency signal is subjected to multi-layer filtering to obtain the target filtered signal, including: In the standard radio frequency signal, the center value of the second preset window is replaced with the median of the second preset window to obtain the first filtered signal; wherein, the size of the second preset window when the liquid sample is in a stable period is larger than the size of the second preset window when the liquid sample is in a sudden change period; The first filtered signal is decomposed into wavelet packets to obtain the second filtered signal; The second filtered signal is smoothed using the Kalman filter algorithm to obtain the target filtered signal.

5. The method according to claim 1, characterized in that, Based on the preset error compensation model, the target filtered signals of different modes are converted into target correction signals, including: Real-time acquisition of liquid sample flow rate data; calculation of signal delay time corresponding to the liquid sample flow rate data based on a preset mapping relationship model; and shifting the target filtered signal at the current moment forward based on the signal delay time to obtain a time-series target correction signal; wherein the sampling frequency of the liquid sample flow rate data is the same as the sampling frequency of the original radio frequency signal; The time-series target correction signal is subjected to frequency domain transformation to identify the bubble signal amplitude; and the time-series target correction signal is subjected to wavelet decomposition, and the low-frequency component representing the effective signal and the high-frequency component representing the bubble interference are determined based on the bubble signal amplitude; the part of the high-frequency component whose amplitude exceeds the preset amplitude is set to 0, and the bubble target correction signal is obtained by inverse wavelet transformation; Based on the thickness of the emulsion layer in the liquid sample and the emulsion layer thickness correction factor, the bubble target correction signal is corrected to obtain the emulsion layer interference correction signal; Based on a preset system error model, the emulsion layer interference correction signal is converted into a target correction signal; the preset system error compensation model is used to calibrate the error of the energy absorption sensor; the preset system error compensation model is a linear transformation model constructed by taking the radio frequency signal of the standard liquid sample as the dependent variable and the radio frequency signal measured by the energy sensor on the standard liquid sample as the independent variable.

6. The method according to claim 1, characterized in that, After the step of weighted summation of the target correction signals of different modalities to obtain the fused target signal, the method further includes: The target signal at the current moment is weighted and summed with the target signal at the previous moment using a second preset weighting coefficient to obtain the target stable signal; wherein the sum of the weights of the target signal at the current moment and the target signal at the previous moment is 1; wherein the weighting coefficient of the target signal at the current moment when the liquid sample is in a stable state is less than the weighting coefficient of the target signal at the current moment when the liquid sample is in an unstable state. Accordingly, the step of determining the average signal strength attenuation rate in each sampling period based on the fused target signal and the number of samples in the horizontal and vertical polarization directions within each sampling period, and obtaining the average signal strength attenuation rate sequence, includes: Based on the target stable signal, the average signal strength attenuation rate within the current sampling period is determined according to the number of samples taken in the horizontal and vertical polarization directions within the current sampling period, thus obtaining an average signal strength attenuation rate sequence.

7. The method according to claim 1 or 6, characterized in that, Based on the fused target signal, the average signal strength attenuation rate in each sampling period is determined according to the number of samples taken in the horizontal and vertical polarization directions within each sampling period, resulting in an average signal strength attenuation rate sequence, including: For each sampling period, the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction are calculated based on the target fused signal in the current sampling period; or, the average signal strength attenuation rate in the horizontal polarization direction and the average vertical signal strength attenuation rate in the vertical polarization direction are calculated based on the target stable signal in the current sampling period. The horizontal polarizability and vertical polarizability are determined based on the number of samples taken in the horizontal polarization direction and the number of samples taken in the vertical polarization direction within the current sampling period. The average signal strength attenuation rate of the current sampling period is obtained by adding the product of the horizontal polarization rate and the average level signal strength attenuation rate, and the product of the vertical polarization rate and the average vertical signal strength attenuation rate. The average signal strength attenuation rate of each sampling period is spliced ​​together to obtain the average signal strength attenuation rate sequence.

8. The method according to claim 1, characterized in that, After the step of converting the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence into liquid sample water content based on a preset conversion relationship to obtain a liquid sample water content sequence, the method further includes: Calculate the sub-mean and sub-standard deviation of the liquid sample water content in each sampling period within the third preset window in the liquid sample water content sequence; If the absolute value of the difference between the liquid sample moisture content and the sub-mean value in the moisture content sequence for m consecutive sampling periods is greater than the preset sub-standard deviation, the moisture content of the current sampling period is replaced with the average moisture content of the previous p sampling periods of the current sampling point; wherein, the preset sub-standard deviation is the product of the sub-standard deviation and the second preset multiple; If the absolute value of the difference between the liquid sample water content and the sub-mean value at n consecutive sampling points in the water content sequence is greater than the preset sub-standard deviation, the health diagnosis of the energy absorption sensor is triggered; where n > m.

9. The method according to claim 1, characterized in that, The energy absorption sensor is an array-type sensor probe, which includes multiple sub-energy absorption sensors, including normal region probes and abnormal region probes. Each sub-energy absorption sensor is configured to acquire raw radio frequency signals of different modes. After preprocessing the raw radio frequency signals of different modes acquired by the energy absorption sensors to obtain the target filtered signal, the method further includes: For each mode of the target filtered signal in each of the sub-energy absorption sensors, calculate the deviation rate between the target filtered signal and the array average signal corresponding to the target filtered signal; If the deviation rate of k consecutive sampling points is greater than or equal to a preset deviation, the sub-energy absorption sensor corresponding to the target filtered signal is determined to be an abnormal region probe. Different weighting coefficients are assigned to the normal region probe and the abnormal region probe respectively, and spatial fusion is performed in the same dimension to obtain the target fusion filtered signal.

10. A liquid sample moisture content detection device based on an energy absorption sensor, characterized in that, The liquid sample moisture content detection device based on the energy absorption sensor includes: The signal preprocessing module is used to preprocess the raw radio frequency signals of different modes acquired by the energy absorption sensor to obtain the target filtered signal; wherein, the preprocessing includes data cleaning and filtering of the raw radio frequency signals; the raw radio frequency signals of different modes include horizontal polarization direction and vertical polarization direction in the same sampling period, and the polarization direction of the raw radio frequency signals of different modes is the same at the same sampling point; An error correction module is used to convert target filtered signals of different modes into target corrected signals according to a preset error compensation model; wherein the preset error compensation model includes at least one of the following: flow velocity-response hysteresis compensation, bubble interference compensation, emulsion layer interference compensation, and sensor calibration error compensation. The signal fusion module is used to perform weighted summation of target correction signals of different modalities to obtain a fused target signal; wherein the sum of the weights corresponding to the target correction signals of different dimensions is 1; The signal strength attenuation rate calculation module is used to determine the average signal strength attenuation rate in each sampling period based on the fused target signal and the number of samples in the horizontal polarization direction and the vertical polarization direction in each sampling period, so as to obtain the average signal strength attenuation rate sequence. The moisture content conversion module is used to convert the average signal intensity attenuation rate in the average signal intensity attenuation rate sequence into the liquid sample moisture content based on a preset conversion relationship, so as to obtain the liquid sample moisture content sequence.

Images