Crude oil moisture content detection method and system based on multi-sensor data fusion

Abstract

The invention discloses a crude oil moisture content detection method and system based on multi-sensor data fusion. The method comprises the following steps: acquiring a crude oil moisture content information data set according to a capacitance sensor, a microwave sensor, a conductivity sensor and a pressure sensor, and preprocessing; performing autocorrelation and partial autocorrelation analysis on the preprocessed data sequence to obtain an optimal hierarchy and order; calculating the information data by using an ARIMA time sequence moisture content residual model to obtain an initial predicted value; calculating a water content residual error data set; taking the water content residual data set as the input of a LightGBM model, and performing information data calculation by adopting the LightGBM model to obtain a further predicted value; and adding the further predicted value and the initial predicted value to obtain the final water content of the crude oil. According to the method, the ARIMA-LightGBM hybrid model is used for fusing the measurement data of the four sensors, so that the detection range of the water content of the crude oil is expanded, and the precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of crude oil water content detection, and in particular to a crude oil water content detection method and system based on multi-sensor data fusion. Background Art

[0002] Petroleum is an important energy source and high-quality chemical raw material, a strategic resource crucial to national livelihoods. During crude oil extraction, to determine the crude oil yield of each well and understand changes in formation oil and gas content and geological structure, it is necessary to continuously measure the volume ratios of the various components in the well output fluid and provide real-time measurement data to optimize production parameters and increase recovery. Water cut is a key monitoring parameter in oil well production processes, playing a crucial role in improving the economic benefits of oil production.

[0003] Since the 1980s, various online methods for measuring crude oil water content have been used in oilfields, primarily capacitance, conductivity, microwave, density, and gamma-ray methods. The capacitance method is accurate within a water content range of less than 30%. The conductivity method achieves an accuracy of ±3% at water content between 50% and 100%. Microwaves have strong penetrating power and high measurement accuracy, achieving microwave water holdup accuracy within 2% when the water content is above 50% or below 20%. The density method has a wide measurement range of 0-100%. However, due to limitations in process and technology, the accuracy, stability, and reliability of these commonly used methods for measuring crude oil water content remain problematic.

[0004] In view of this, this application is hereby filed. Summary of the Invention

[0005] The purpose of the present invention is to provide a crude oil water content detection method and system based on multi-sensor data fusion, which can achieve high-precision and full-range water content measurement of crude oil, and provide important water content reference value for oil well production.

[0006] The present invention is achieved through the following technical solutions:

[0007] In a first aspect, the present invention provides a method for detecting the water content of crude oil based on multi-sensor data fusion, the method comprising:

[0008] The crude oil water content information data set (X1, X2, X3, X4) is collected using capacitance sensors, microwave sensors, conductivity sensors, and pressure sensors, and the information data set is preprocessed to obtain a water content preprocessed data sequence;

[0009] The optimal stratum and order are obtained by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessing data series. Based on the optimal stratum and order, an ARIMA time series moisture content residual model is constructed. The ARIMA time series moisture content residual model is used to calculate the information data to obtain the initial prediction value. The moisture content residual data set is then calculated based on the moisture content preprocessing data series.

[0010] The water cut residual dataset is used as the input of the LightGBM model. The LightGBM model is constructed and trained to perform information data calculations to obtain the residual prediction sequence and further prediction values. The further prediction values ​​are added to the initial prediction values ​​to obtain the final crude oil water cut.

[0011] Furthermore, the capacitance sensor, microwave sensor, conductivity sensor and pressure sensor correspond to the capacitance method, microwave method, conductivity method and density method of the traditional crude oil water content measurement methods respectively.

[0012] Furthermore, by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessed data series, the optimal hierarchy and order are obtained, including:

[0013] The unit root test (ADF) method is used to determine the stationarity of the moisture content preprocessed data series and obtain a stationary time series;

[0014] The optimal hierarchy and order are obtained by performing autocorrelation and partial autocorrelation analysis on the stationary time series.

[0015] Furthermore, the formulas for autocorrelation and partial autocorrelation analysis are:

[0016]

[0017]

[0018] Where ρi is the autocorrelation coefficient, Cov(·) is the autocovariance function, Var(·) is the variance function, and α ij is the partial correlation coefficient.

[0019] Furthermore, the expression of the ARIMA time series moisture content residual model is:

[0020]

[0021] Among them, Ψt is the stationary series after difference, μ is the constant term, p is the autoregressive order, q is the moving average term, ε t It's wandering white noise. is the AR autoregressive model, It is the MA moving average model.

[0022] Furthermore, the Bayesian criterion (BIC) coefficient is used to construct the size of the residual model of the moisture content of the ARIMA time series.

[0023] Furthermore, the LightGBM model is trained, including:

[0024] The LightGBM model trains a basic learner f(x). The next learner will perform more learning on data samples with large prediction deviations from the previous learner. The iterative training obtains the crude oil water content output and serves as the residual prediction sequence.

[0025] Furthermore, the LightGBM model is expressed as a weighted sum function of multiple base learners. The expression used to learn and train the LightGBM model is:

[0026]

[0027] in, is the weight of the nth base learner, x is the input data, f n (x) is the output of the nth base learner, F n (x) is the total output of the LightGBM model.

[0028] Furthermore, training the LightGBM model also includes:

[0029] The GOSS algorithm and EFB algorithm are used to discretize the training samples of the moisture content residual data set and perform feature extraction.

[0030] In a second aspect, the present invention further provides a crude oil water content detection system based on multi-sensor data fusion, which uses the above-mentioned crude oil water content detection method based on multi-sensor data fusion; the system comprises:

[0031] An acquisition and preprocessing unit is used to collect a crude oil water content information data set (X1, X2, X3, X4) using a capacitance sensor, a microwave sensor, a conductivity sensor, and a pressure sensor, and preprocess the information data set to obtain a water content preprocessing data sequence;

[0032] The initial prediction value calculation unit is used to obtain the optimal hierarchy and order by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessing data sequence; based on the optimal hierarchy and order, an ARIMA time series moisture content residual model is constructed; the information data is calculated using the ARIMA time series moisture content residual model to obtain the initial prediction value; and the moisture content residual data set is calculated based on the moisture content preprocessing data sequence;

[0033] The final crude oil water content calculation unit is used to take the water content residual data set as the input of the LightGBM model, build and train the LightGBM model to perform information data calculation, obtain the residual prediction sequence and further prediction value; and add the further prediction value to the initial prediction value to obtain the final crude oil water content.

[0034] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0035] The present invention provides a crude oil water content detection method and system based on multi-sensor data fusion. The present invention fuses measurement data sets of capacitance sensors, conductivity sensors, microwave sensors and pressure sensors to achieve accurate measurement of the entire range of crude oil water content. The data fusion adopts a hybrid model combining the RIMA time series water content residual model and the LightBGM model to optimize and quickly process the measurement data of the four sensors. The reliability and accuracy of the crude oil water content measurement model are improved, the scope of crude oil water content detection is expanded and the accuracy is improved, which better meets the current demand for crude oil water content measurement technology in oil production. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0037] Figure 1 This is a flow chart of a method for detecting water content in crude oil based on multi-sensor data fusion according to the present invention;

[0038] Figure 2 This is a detailed flow chart of a method for detecting water content in crude oil based on multi-sensor data fusion according to the present invention;

[0039] Figure 3 This is a flow chart of the ARIMA time series moisture content residual model of the present invention;

[0040] Figure 4 This is a flow chart of the LightBGM model algorithm of the present invention;

[0041] Figure 5 The test results of the capacitance method, density method, microwave method, conductivity method and the ARIMA-Light fusion method of the present invention are shown in FIG.

[0042] Figure 6 This is a structural block diagram of a crude oil water content detection system based on multi-sensor data fusion in the present invention. DETAILED DESCRIPTION

[0043] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0044] Existing methods for online measurement of crude oil moisture content primarily include capacitance, conductivity, microwave, density, and gamma-ray methods. The capacitance method is accurate within a moisture content range of less than 30%. The conductivity method achieves a measurement accuracy of ±3% within a moisture content range of 50%-100%. Microwaves have strong penetrating power and high measurement accuracy, achieving a microwave water holdup accuracy of less than 2% when the moisture content is above 50% or below 20%. The density method has a wide measurement range of 0-100%. Due to limitations in process and technology, these commonly used methods for measuring crude oil moisture content still face numerous challenges in accuracy, stability, and reliability. Therefore, in response to the unique characteristics of crude oil production in my country, the present invention considers combining multiple sensors to measure crude oil moisture content and improve detection accuracy.

[0045] Multi-sensor data fusion is a new data processing method. Information fusion involves the integrated processing of information from multiple sensors or sources to produce a more accurate and reliable picture. For example, when the moisture content is above 30%, the capacitance method's accuracy is significantly reduced, while when the moisture content is below 30%, the conductivity method's accuracy is significantly reduced. Multi-sensor data fusion combines these two methods, expanding the measurement range and improving accuracy. Currently, multi-sensor data fusion methods can be categorized into two types: probabilistic and statistical methods and artificial intelligence methods. Probabilistic and statistical methods include estimation theory, Bayesian methods, and decision statistics theory; artificial intelligence methods include evidence theory, logical reasoning, and artificial neural networks.

[0046] In recent years, with the continuous research of many scholars, it has been found that sometimes a single model is not thorough enough in mining the information of the research object, and the effect is not stable enough. At this time, the research has turned to the hybrid model. The Autoregressive Integrated Moving Average (ARIMA) model is a model that converts a non-stationary time series into a stationary time series, and then regresses the dependent variable only on its lagged value and the present value and lagged value of the random error term. LightGBM is a distributed gradient boosting (GBDT) framework based on the decision tree algorithm open sourced by the Distributed Machine Learning Toolkit Team of Microsoft Research Asia. It supports efficient parallel training and has the advantages of faster training speed, lower memory consumption, better accuracy, support for distributed processing of massive data, etc. The present invention proposes an ARIMA-LightGBM hybrid model, which has a good fusion processing effect on the data generated by a variety of sensors for traditional measurement of crude oil water content. The novel multi-sensor data fusion method provides ideas for new sensors required for crude oil water content measurement, which has important social and economic significance.

[0047] The present invention provides a method for detecting the water content of crude oil based on multi-sensor data fusion. The water content of crude oil is first measured using a capacitance method, a conductivity method, a microwave method, and a density method. The measurement results of the four methods are then fused through an ARIMA-LightGBM hybrid model to obtain the water content of crude oil. The time series ARIMA model is used to mine the linear characteristics of the data, and the LightGBM regression model is used to mine the nonlinear characteristics of the data. The hybrid model combining the two has both linear and nonlinear processing capabilities. Among the traditional methods for measuring the water content of crude oil, the capacitance method is suitable for crude oil with low water content, the conductivity method and the microwave method are suitable for crude oil with high water content, and the density method is suitable for crude oil with a full range of water content but the accuracy is not high. The present invention uses the ARIMA-LightGBM hybrid model to fuse the measurement data of the four sensors, which expands the scope of crude oil water content detection and improves the accuracy.

[0048] Example 1

[0049] like Figure 1 and Figure 2 As shown, the present invention provides a method for detecting the water content of crude oil based on multi-sensor data fusion, the method comprising:

[0050] Step 1: Collect the crude oil water content information dataset (X1, X2, X3) using capacitance sensor, microwave sensor, conductivity sensor and pressure sensor. 3, X4), and preprocess the information data set to obtain a moisture content preprocessing data sequence;

[0051] The four sensors in step 1 (capacitive sensor, microwave sensor, conductivity sensor and pressure sensor) correspond to the capacitance method, microwave method, conductivity method and density method of traditional crude oil water content measurement methods respectively.

[0052] This is because, considering the measurement range, the capacitance method suitable for low water content measurement, the conductivity method suitable for high water content measurement, and the density method suitable for the entire range but with low accuracy are combined to improve the accuracy of crude oil water content in the entire range; and the measurement method selects the capacitance method, conductivity method, density method and other contact measurement methods combined with the microwave non-contact measurement method to compensate for the influence of mineralization on accuracy.

[0053] In this embodiment, the crude oil water content samples were prepared from diesel, tap water, and an emulsifier to form 11 crude oil samples with water contents of 0, 10%, 20%, ..., and 100%. 120 ml of each water content level was prepared and tested at various temperatures.

[0054] The above four sensor data collection methods are all known methods, and the collected data sets are all voltage data. The capacitance sensor causes the capacitance to change according to the measured water content, and then the voltage value changes after detection; the microwave sensor can replace the more difficult microwave power measurement by measuring the voltage amplitude; the magnitude of the induced voltage of the conductivity sensor is inversely proportional to the conductivity of the fluid flowing through the sensor; the pressure sensor can reflect the crude oil density and thus the water content by calculating the pressure difference.

[0055] The steps of preprocessing the collected data include outlier processing and missing value processing.

[0056] Specifically, the water content of crude oil samples was tested using capacitance, density, microwave and conductivity methods according to known methods. The test results of 30 samples are as follows: Figure 5 shown.

[0057] Step 2: Perform autocorrelation and partial autocorrelation analysis on the moisture content preprocessing data sequence to obtain the optimal hierarchy and order; construct an ARIMA time series moisture content residual model based on the optimal hierarchy and order; calculate the information data using the ARIMA time series moisture content residual model to obtain an initial prediction value; and calculate the moisture content residual data set based on the moisture content preprocessing data sequence;

[0058] Specifically, the expression of the ARIMA time series moisture content residual model is:

[0059]

[0060] Among them, Ψt is the stationary series after difference, μ is the constant term, p is the autoregressive order, q is the moving average term, ε t It's wandering white noise. is the AR autoregressive model, It is the MA moving average model.

[0061] Specifically, in step 2, the optimal hierarchy and order are obtained by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessed data sequence, including:

[0062] Step A, using the unit root test (ADF) method to determine the stationarity of the moisture content preprocessed data sequence, and obtain a stationary time series; specifically, the stationarity of the moisture content preprocessed data sequence is determined by the unit root test (ADF). If the obtained statistic is significantly lower than the critical statistics of three confidence levels (1%, 5%, and 10%), the rejection hypothesis of the original sequence is that the sequence is stationary, and the sequence is stabilized by a finite-order difference operation.

[0063] Step B, obtain the optimal hierarchy and order by performing autocorrelation and partial autocorrelation analysis on the stationary time series.

[0064] The optimal stratum and order are determined by calculating the sample autocorrelation function and partial autocorrelation function through the observed values ​​of the moisture content sample.

[0065] Specifically, the formulas for autocorrelation and partial autocorrelation analysis in step 2 are:

[0066]

[0067]

[0068] Where ρi is the autocorrelation coefficient, Cov(·) is the autocovariance function, Var(·) is the variance function, and α ij is the partial correlation coefficient.

[0069] Specifically, when training the ARIMA time series moisture content residual model, increasing the values ​​of parameters p and q will increase the complexity of the model and cause overfitting. Therefore, the present invention uses the Bayesian Criterion (BIC) coefficient to construct the size of the ARIMA time series moisture content residual model. The BIC measurement standard formula (4) is expressed as follows:

[0070] BIC=kln(n)-2ln(L) (4)

[0071] Specifically, the flowchart of constructing the ARIMA time series moisture content residual model in step 2 is as follows: Figure 3 As shown in the figure, the water content residual data is tested with white noise, and it is determined that the established model can effectively extract the information of the original data.

[0072] Step 3: Use the water content residual dataset as the input of the LightGBM model, build and train the LightGBM model to perform information data calculation, and obtain the residual prediction sequence and further prediction value; and add the further prediction value to the initial prediction value to obtain the final crude oil water content.

[0073] Specifically, training the LightGBM model in step 3 includes:

[0074] The LightGBM model trains a basic learner f(x). The next learner will perform more learning on data samples with large prediction deviations from the previous learner. The iterative training obtains the crude oil water content output and serves as the residual prediction sequence.

[0075] In this embodiment, the algorithm flow chart of the LightBGM model is as follows: Figure 4 As shown, the LightGBM model is expressed as a weighted sum function of multi-base learners. The expression used to learn and train the LightGBM model is:

[0076]

[0077] in, is the weight of the nth base learner, x is the input data, f n (x) is the output of the nth base learner, F n (x) is the total output of the LightGBM model.

[0078] In this embodiment, training the LightGBM model also includes:

[0079] The gradient-based single-side sampling algorithm (GOSS) and the mutually exclusive feature bundling algorithm (EFB) are used to discretize and extract features from the training samples of the moisture content residual dataset. Specifically, the moisture content training set data is first sorted according to gradient, and a preset ratio is set to retain data samples with gradients above this ratio among all samples. Data samples with gradients below this ratio are not directly discarded. Instead, a sampling ratio is set and samples are proportionally extracted from samples with smaller gradients. When calculating information gain, the data sets with smaller gradients are multiplied by a coefficient to amplify them. Specifically, the EFB algorithm extracts features from the moisture content training samples by adding an offset constant to different features within the bundled feature cluster, so that the values ​​of different features are distributed over different value spaces.

[0080] Parameters determine the quality of the trained model. Selecting an appropriate maximum tree depth, minimum number of leaf records, and learner weight reduction coefficient is crucial. Specifically, the decision tree leaf growth strategy selects the most profitable node and sets a maximum leaf depth limit to prevent overfitting caused by excessive leaf depth.

[0081] A histogram-based partition point search algorithm is used to discretize the eigenvalues ​​into several value spaces, reducing the algorithm complexity of finding the optimal splitting point, effectively reducing the computing and storage costs, and speeding up model training.

[0082] The final result of the crude oil water content (i.e., the final crude oil water content) is the sum of the initial prediction value of the ARIMA model and the prediction value of the LightGBM model. The test results of 30 crude oil samples are as follows: Figure 5 As shown in the figure, the prediction effect is better than that of the four traditional moisture content methods.

[0083] The present invention fuses the measurement data sets of capacitance sensors, conductivity sensors, microwave sensors and pressure sensors to achieve accurate measurement of the full range of crude oil water content. Data fusion adopts a hybrid model combining the RIMA time series water content residual model and the LightBGM model, which can optimize and quickly process the measurement data of the four sensors. The reliability and accuracy of the crude oil water content measurement model are improved, which better meets the current demand for crude oil water content measurement technology in oil production.

[0084] Example 2

[0085] like Figure 6 As shown, the difference between this embodiment and embodiment 1 is that this embodiment provides a crude oil water content detection system based on multi-sensor data fusion, which uses the crude oil water content detection method based on multi-sensor data fusion in embodiment 1; the system includes:

[0086] An acquisition and preprocessing unit is used to collect a crude oil water content information data set (X1, X2, X3, X4) using a capacitance sensor, a microwave sensor, a conductivity sensor, and a pressure sensor, and preprocess the information data set to obtain a water content preprocessing data sequence;

[0087] The initial prediction value calculation unit is used to obtain the optimal hierarchy and order by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessing data sequence; based on the optimal hierarchy and order, an ARIMA time series moisture content residual model is constructed; the information data is calculated using the ARIMA time series moisture content residual model to obtain the initial prediction value; and the moisture content residual data set is calculated based on the moisture content preprocessing data sequence;

[0088] The final crude oil water content calculation unit is used to take the water content residual data set as the input of the LightGBM model, build and train the LightGBM model to perform information data calculation, obtain the residual prediction sequence and further prediction value; and add the further prediction value to the initial prediction value to obtain the final crude oil water content.

[0089] The execution process of each unit can be performed according to the process steps of a crude oil water content detection method based on multi-sensor data fusion in Example 1, and will not be described in detail in this embodiment.

[0090] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0091] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0094] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting water content in crude oil based on multi-sensor data fusion, characterized in that: The method includes: A crude oil water content information data set is collected using a capacitance sensor, a microwave sensor, a conductivity sensor, and a pressure sensor, and the information data set is preprocessed to obtain a water content preprocessed data sequence; The optimal stratum and order are obtained by performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessing data sequence; an ARIMA time series moisture content residual model is constructed based on the optimal stratum and order; the information data is calculated using the ARIMA time series moisture content residual model to obtain an initial prediction value; and a moisture content residual data set is calculated based on the moisture content preprocessing data sequence; The water content residual data set is used as the input of the LightGBM model, and the LightGBM model is constructed and trained to perform information data calculation to obtain a residual prediction sequence and a further prediction value; and the further prediction value is added to the initial prediction value to obtain the final crude oil water content.

2. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 1, characterized in that: The capacitance sensor, microwave sensor, conductivity sensor and pressure sensor respectively correspond to the capacitance method, microwave method, conductivity method and density method of traditional crude oil water content measurement methods.

3. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 1, characterized in that: By performing autocorrelation and partial autocorrelation analysis on the moisture content preprocessed data series, the optimal hierarchy and order are obtained, including: Using the unit root test method to determine the stationarity of the moisture content preprocessed data sequence, and obtain a stationary time series; The optimal hierarchy and order are obtained by performing autocorrelation and partial autocorrelation analysis on the stationary time series.

4. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 1 or 3, characterized in that: The formulas for the autocorrelation and partial autocorrelation analysis are: Where ρi is the autocorrelation coefficient, Cov(·) is the autocovariance function, Var(·) is the variance function, and α ij is the partial correlation coefficient.

5. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 1, characterized in that: The expression of the ARIMA time series moisture content residual model is: Among them, Ψt is the stationary series after difference, μ is the constant term, p is the autoregressive order, q is the moving average term, ε t It's wandering white noise. is the AR autoregressive model, It is the MA moving average model.

6. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 5, characterized in that: The Bayesian criterion coefficient is used to construct the size of the ARIMA time series moisture content residual model.

7. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 1, characterized in that: Training the LightGBM model includes: The LightGBM model trains a basic learner f(x). The next learner will perform more learning on data samples with large prediction deviations from the previous learner. The iterative training obtains the crude oil water content output and serves as the residual prediction sequence.

8. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 7, characterized in that: The LightGBM model is expressed as a weighted sum function of multi-base learners. The expression used to learn and train the LightGBM model is: in, is the weight of the nth base learner, x is the input data, f n (x) is the output of the nth base learner, F n (x) is the total output of the LightGBM model.

9. The method for detecting water content in crude oil based on multi-sensor data fusion according to claim 7, characterized in that: Training the LightGBM model also includes: The GOSS algorithm and the EFB algorithm are used to perform discretization sampling and feature extraction on the training samples of the moisture content residual data set.

10. A crude oil water content detection system based on multi-sensor data fusion, characterized in that: The system uses a crude oil water content detection method based on multi-sensor data fusion as described in any one of claims 1 to 9; the system includes: an acquisition and preprocessing unit, configured to collect a crude oil water content information data set using a capacitance sensor, a microwave sensor, a conductivity sensor, and a pressure sensor, and preprocess the information data set to obtain a water content preprocessing data sequence; An initial prediction value calculation unit is used to obtain an optimal hierarchy and order by performing autocorrelation and partial autocorrelation analysis on a moisture content preprocessing data sequence; construct an ARIMA time series moisture content residual model based on the optimal hierarchy and order; calculate information data using the ARIMA time series moisture content residual model to obtain an initial prediction value; and calculate a moisture content residual data set based on the moisture content preprocessing data sequence; The final crude oil water content calculation unit is used to use the water content residual data set as the input of the LightGBM model, build and train the LightGBM model to perform information data calculation, obtain the residual prediction sequence and further prediction value; and add the further prediction value to the initial prediction value to obtain the final crude oil water content.

Images