CN116822999A - Method and system for predicting monitoring density of oil product of oil mixing interface of finished oil pipeline - Google Patents

Abstract

The invention belongs to the field of data processing technology. In order to solve the problem of being unable to reliably predict the density of subsequent oil product monitoring, a method and system for predicting the density of subsequent oil product monitoring at the mixed interface of a refined oil pipeline is proposed. The refined oil pipeline selected from historical data The hydraulic and thermal information of the upstream and downstream stations, the monitoring density of the upstream and downstream stations for oil products in the preceding row, and the monitoring density difference between the upstream and downstream stations of the preceding row of oil products are used as key input feature variables. The difference in monitoring density between the upstream and downstream stations of the trailing oil products is selected as For output variables, the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables. Based on the modal identification results, the maximum expectation algorithm is used for training to obtain the monitoring density difference between upstream and downstream oil products corresponding to the modal. value correction model; and combined with the obtained monitoring density value of the oil product upstream station behind the product oil pipeline, the prediction results are obtained. The density of oil monitoring can be accurately predicted after the mixed oil interface.

Description

Method and system for density prediction of oil product monitoring at the mixed oil interface of refined oil pipelines

Technical field

The invention belongs to the field of data processing technology, and in particular relates to a density prediction method and system for oil product monitoring after the mixed oil interface of a product oil pipeline.

Background technique

The statements in this section merely provide background technical information related to the present invention and do not necessarily constitute prior art.

For economic reasons, multiple refined oil products are usually transported continuously in the same refined oil pipeline in a certain batch sequence. During the pipeline transportation process, there will be an oil-mixed interface between adjacent batches. Although it is difficult for the density meter installed at the station to obtain the true density value of the oil products before and after the oil-mixed interface, its real-time feedback of the oil-mixed interface monitors the density changes over time. It is the core data for operators to handle the mixed oil interface. Specifically, on-site, the monitoring value of the mixed oil interface density is generally converted into a time change curve of the mixed oil concentration distribution. When the mixed oil concentration value drops to a certain threshold, the batch cutting method is used to process the mixed oil interface. However, the density meter can only sense the current density monitoring value of the mixed oil interface at the station, and cannot predict the density monitoring value of the pure oil that has not yet arrived at the station. If the oil density monitoring value can be accurately controlled, key data support can be provided to accurately guide on-site oil batch cutting work. However, the hydraulic and thermal conditions in the pipeline transportation process are complex and changeable. Based on the existing oil density calculation formula, it is impossible to accurately obtain the monitoring density of downstream oil products. Hardware equipment generally faces zero-point drift, resulting in the monitoring density values of the same oil products at upstream and downstream stations. There are obvious differences that trouble the oil batch cutting work. In addition, frequent changes in pipeline transportation conditions lead to multi-modal characteristics in the data. If the data-driven modeling method is directly used to predict the density of downstream oil products, the model will easily fall into the error of overfitting and cannot provide reliable predictions of the monitoring density of downstream oil products. result.

Contents of the invention

In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for predicting the oil product density after the mixed interface of a product oil pipeline. The Gaussian mixture regression algorithm is used to perform multi-modal identification of the product oil pipeline data and utilize the maximum expectation. The algorithm is trained based on multi-modal recognition results to obtain a correction model for predicting the monitoring density difference of upstream and downstream oil products at upstream and downstream stations, so that the monitoring density of downstream oil products at the mixed oil interface can be accurately predicted.

In order to achieve the above object, the first aspect of the present invention provides a density prediction method for oil product monitoring after the oil mixed interface of a product oil pipeline, including:

Obtain historical data of the mixed oil interface of product oil pipelines;

Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream and downstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the upstream and downstream stations of the subsequent oil products. The station monitoring density difference is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained;

According to the mode identification results, the maximum expectation algorithm is used for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the mode;

The correction model of the density difference between upstream and downstream oil product monitoring stations is used to predict the mixed oil interface of the product oil pipeline to be predicted. Combined with the obtained monitoring density value of the oil product upstream station for the product oil pipeline, the product oil pipeline to be predicted is obtained. Monitoring density value of downstream oil products at downstream stations.

The second aspect of the invention provides an oil product monitoring and density prediction system behind the mixed interface of a refined oil pipeline, including:

Acquisition module: obtain historical data of the mixed oil interface of the refined oil pipeline;

Gaussian module: Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the following oil products The difference in monitoring density between the upstream and downstream stations of the product is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained;

Training module: Based on the modal identification results, use the maximum expectation algorithm for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the modal;

Prediction module: Use the correction model of the monitoring density difference between the upstream and downstream stations of the downstream oil products to predict the mixed oil interface of the refined oil pipeline to be predicted, and combine it with the obtained monitoring density value of the upstream oil products in the refined oil pipeline to obtain the prediction Monitoring density value of downstream oil products at downstream stations of refined oil pipelines.

One or more of the above technical solutions have the following beneficial effects:

In the present invention, the historical data of the refined oil pipeline is obtained, the hydraulic and thermal information of the upstream and downstream stations is selected from the historical data, and the monitoring density of the upstream and downstream stations of the preceding oil products and the difference in monitoring density of the upstream and downstream stations of the preceding oil products are used as the key Input the characteristic variables, and use the monitoring density difference between the upstream and downstream stations of the downstream oil products as the output variable. The Gaussian regression algorithm is used to identify the mode. The modal recognition results are trained based on the maximum expectation algorithm, and the upstream and downstream stations of the downstream oil products are obtained. The on-site monitoring density difference correction model uses the obtained upstream and downstream oil product monitoring density difference correction model to predict the to-be-predicted on-offline oil product upstream and downstream station monitoring density difference; a Gaussian regression algorithm is used to perform multiple The identification of modal characteristics improves the accuracy of oil monitoring density prediction after the mixed oil interface, and is of great significance in guiding on-site oil batch management.

Additional advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Description of the drawings

The description and drawings that constitute a part of the present invention are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

Figure 1 is a schematic diagram of the migration measured at the oil-mixed interface in Embodiment 1 of the present invention;

Figure 2 is a common mixed oil density distribution curve in Embodiment 1 of the present invention;

Figure 3 is a diagram showing the distortion of the mixed oil concentration distribution curve caused by the density prediction deviation of subsequent oil monitoring in Embodiment 1 of the present invention;

Figure 4 is a flow chart of the subsequent oil product density monitoring value prediction model in Embodiment 1 of the present invention;

Figure 5 is a schematic diagram of the correction model for the density monitoring difference between upstream and downstream oil product stations in Embodiment 1 of the present invention.

Detailed ways

It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terms used herein are for the purpose of describing specific embodiments only, and are not intended to limit the exemplary embodiments according to the present invention.

The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.

Embodiment 1

This embodiment discloses a density prediction method for oil product monitoring behind the mixed oil interface of a product oil pipeline, including:

Obtain historical data of the mixed oil interface of product oil pipelines;

Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream and downstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the upstream and downstream stations of the subsequent oil products. The station monitoring density difference is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained;

According to the mode identification results, the maximum expectation algorithm is used for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the mode;

The correction model of the density difference between upstream and downstream oil product monitoring stations is used to predict the mixed oil interface of the product oil pipeline to be predicted. Combined with the obtained monitoring density value of the oil product upstream station for the product oil pipeline, the product oil pipeline to be predicted is obtained. Monitoring density value of downstream oil products at downstream stations.

The method for predicting the density of oil products monitored at the mixed oil interface of the refined oil pipeline provided by this embodiment mainly includes three parts: establishing a mixed oil database at the upstream and downstream stations of the refined oil pipeline, and establishing the density difference correction for monitoring the density of the upstream and downstream oil products at the upstream and downstream stations. Model, oil product monitoring density prediction after mixed oil interface.

(1) Establish a mixed oil database at upstream and downstream stations of refined oil pipelines

The establishment of a mixed oil database at the upstream and downstream stations of the refined oil pipeline is based on the measurement equipment at the upstream and downstream stations. It obtains the pipeline water and thermal information and the front and rear oil product monitoring densities and establishes a database for subsequent establishment of the rear oil product monitoring density difference correction. The model provides data support. This includes determining the outlet temperature and pressure of the upstream station, the inlet temperature and pressure of the downstream station, the density of oil product monitoring before and after the upstream station, and the density of oil product monitoring before and after the downstream station within a certain period of time.

Determining the temperature and pressure at the outlet of the upstream station is specifically: collecting the temperature and pressure at the outlet of the upstream station of the refined oil pipeline within a certain period of time, where the temperature monitoring sample size is , the pressure monitoring sample size is/> , after average processing, the hourly average temperature of the upstream station outlet can be obtained/> , ℃; time average pressure/> , unit: MPa:

(1)

(2)

in, Collected for the upstream station/> temperature data points, ℃;/> Collected for the upstream station/> pressure data points, MPa.

Determining the temperature and pressure at the downstream station inlet is specifically: collecting the temperature and pressure at the downstream station inlet of the refined oil pipeline within a certain period of time, where the temperature monitoring sample size is , the pressure monitoring sample size is/> , after average processing, the hourly average temperature of the downstream station inlet can be obtained/> , unit: ℃; hourly average pressure/> , unit MPa:

(3)

(4)

in, Collected for downstream stations/> temperature data points, ℃,/> Collected for downstream stations/> pressure data points, MPa.

Determine the density difference of forward oil monitoring obtained from upstream and downstream stations Density difference from subsequent oil monitoring Specifically: based on the upstream and downstream station density meters, obtain the forward oil density monitoring value of the upstream station/> , unit: kg/m3, oil density monitoring value behind the upstream station/> , unit: kg/m3; oil density monitoring value in front of downstream station/> , unit: kg/m3; oil density monitoring value behind the downstream station/> , unit: kg/m3; based on this, calculate the density difference of upstream and downstream oil product monitoring/> Difference in density from upstream and downstream oil product monitoring/> :

(5)

(6)

(2) Establish a density difference correction model for upstream and downstream oil product monitoring stations

The establishment of the density difference correction model for upstream and downstream oil product monitoring at upstream and downstream stations is based on the historical data of the existing pipeline mixed oil interface, combined with the Gaussian Mixture Regression Model (GMR), to form the density difference for downstream oil product monitoring. Modify the model, specifically including:

Select key feature variables for modal recognition: let and/> Respectively represent the input variable matrix and output variable matrix of mixed oil data,/> is the sample size of the training set,/> For the matrix transpose operation, Characterization/> training set sample input variable vector.

In order to accurately explore the multi-modal information hidden in the pipeline mixed oil data, the temperatures obtained from the upstream and downstream stations were selected. and , pressure/> with/> And the forward oil monitoring density obtained from the upstream station/> , rear oil monitoring density/> And the forward oil monitoring density difference obtained from upstream and downstream stations/> As the key input feature variable for the GMR algorithm to identify multi-modal data.

It is assumed in GMR that the input variable is given by consists of a Gaussian distribution. Input variables are at/> distribution, that is, the /> The expression of the marginal probability density function under each mode/> as follows:

(7)

in, is a Gaussian distribution,/> ,/> Respectively refer to No./> The mean vector and covariance matrix in the Gaussian distribution under each mode.

Determine the key input variables of the regression: the difference in monitoring density of the upstream and downstream oil products obtained from the upstream and downstream stations As a key input variable in the regression process of GMR algorithm/> , due to the difference in density monitored by upstream and downstream oil product stations is the variable to be predicted, and the relationship between the two can be expressed as:

+/> (8)

in, Indicates the location/> Gaussian white noise in each mode has a mean value of 0 and a variance of/> Gaussian distribution, T represents the transpose operation. In GMR,/> Each modality is represented by the corresponding/> A local model representation,/> For the first/> regression coefficients in a local model.

Density difference correction model for oil product monitoring after training: Model parameters can be estimated based on the existing historical data of the mixed oil interface in pipelines and combined with the Expectation Maximization (EM) algorithm. Use latent variables Characterizes the modal belonging of each sample, when/> Time indicates the number/> The training set samples belong to the /> a modal. Calculate/> Belongs to/> Posterior probability of a mode/> , and define statistics/> :

(9)

(10)

in, For the given number/> When inputting and outputting feature information of a sample, the Samples belong to/> Conditional probability of a mode;/> Characterization/> Samples belong to/> mode, the conditional probability of the output variable with respect to the input variable;/> For the first/> Samples belong to/> mode, the conditional probability distribution of the input variable.

Based on the obtained posterior probability , can calculate the Gaussian distribution weight of each mode/> ,mean/> and covariance/> , the expression is as follows:

(11)

(12)

(13)

(14)

(15)

in, ;/> is a diagonal matrix , /> .

The density prediction of the downstream oil product monitoring at the mixed oil interface of the refined oil pipeline is to predict the mixed oil interface where the downstream oil product monitoring density is to be predicted. Based on the hydrothermal information and oil product monitoring density information obtained by the upstream and downstream station measurement equipment, the prediction of the downstream oil product monitoring density is Oil monitoring density difference , combined with the downstream oil monitoring density value obtained from the upstream station/> , predict the monitoring density value of downstream oil products/> , specifically including:

Determine the key variable information of the sample to be predicted: Take the following Take a sample to be predicted as an example, and combine the upstream and downstream measurement equipment to obtain the temperature of the upstream and downstream stations/> with/> , pressure/> with/> And the forward oil monitoring density obtained from the upstream station/> , rear oil monitoring density/> And the obtained density difference between the upstream and downstream stations of Qianxing Oil Products/> , which constitutes the modal identification variable in the GMR algorithm/> , monitor the density difference between the upstream and downstream stations of Qianxing Oil/> Regressors that make up the GMR algorithm .

Oil product monitoring density difference after prediction : For the first/> samples to be predicted/> ,/> is a known quantity, is the quantity to be predicted, its distribution can be calculated based on equation (16)/> Posterior probability of a mode/> :

(16)

in, For the given number/> When inputting and outputting feature information of a sample, the Samples belong to/> The conditional probability of a mode,/> For the first/> Samples belong to/> The probability of a mode,/> For the first/> Samples belong to/> mode, the conditional probability distribution of the input variable.

Correspondingly, the quantity to be predicted can be obtained Conditional probability expression/> :

(17)

in, Given input variables, the quantity to be predicted/> conditional probability expression; For output variables/> Obey the mean as/> , the variance is/> Gaussian distribution.

Consider expectations as predicted results/> , so the subsequent oil monitoring density difference/> Predicted value/> The expression is:

(18)

in, for/> In given/> Conditional expectations below.

Predict the oil product monitoring density value : Combined with the downstream oil monitoring density value obtained from the upstream station/> , predict the oil product monitoring density value behind the downstream station/> :

(19)

Figure 1 is a schematic diagram of measured migration at the mixed oil interface. Various refined oil products are pumped from upstream manufacturers to downstream users based on refined oil pipelines in certain batches. During the passing process of the mixed oil interface, accurately estimating the subsequent oil monitoring density and determining the mixed oil concentration distribution information is a key prerequisite for achieving accurate oil batch cutting and improving the economic benefits of the refined oil pipeline network operation.

Figure 2 shows a common mixed oil density distribution curve. When the mixed oil interface passes through the upstream station, the density sensor installed at the station first detects the density timing information of the preceding oil product in the batch sequence, and then gradually senses it. After receiving the oil density monitoring information. The mixed oil concentration distribution information needs to be calculated based on the preceding and following oil product monitoring density information. However, when the mixed oil interface migrates in the pipeline between two adjacent stations, the downstream station can only sense the preceding oil product monitoring density. At present, there is still a lack of methods for directly monitoring the density of oil products in the pipeline. Therefore, it is necessary to combine on-site data to establish a high-precision prediction method for monitoring the density of oil products in the pipeline to provide data support for oil batch cutting work.

Figure 3 shows the mixed oil concentration distribution information error caused by the subsequent oil product monitoring density prediction error, where δ represents the subsequent oil product monitoring density prediction error, and δ=0 represents the actual mixed oil concentration distribution curve. As the monitoring density prediction error of subsequent oil products gradually increases, the corresponding mixed oil concentration distribution curve deviates more and more from the true curve: the mixed oil concentration distribution should be in the [0,1] interval, but when the subsequent oil products When there is an error in the monitoring density prediction result, the mixed oil concentration value may appear negative, and the calculation result violates the laws of physics. In addition, if the mixed oil concentration value of 1% is defined as the oil batch cutting threshold point, when δ increases from 0 to 1, 3, and 5, the corresponding time deviations of the oil batch cutting threshold points that represent the mixed oil interface are respectively for 216s, 450s and 624s. The above results once again illustrate the importance of accurate prediction of subsequent oil monitoring density for accurate oil batch cutting work.

Figure 4 shows the downstream oil density monitoring value prediction model process. First, based on the measurement equipment installed at upstream and downstream stations, temperature, pressure and oil density information are obtained respectively. It should be noted that for accurate modeling, the upstream station density meter can know the order of oil batches after monitoring the density of oil products in the front and rear rows. For example, gasoline pushes diesel, which is referred to as the gasoline-diesel interface; diesel pushes gasoline, which is referred to as the diesel-gasoline interface. By analogy, a corresponding subsequent oil product density monitoring difference correction model is established accordingly.

Figure 5 shows a schematic diagram of the difference correction model for downstream oil product density monitoring; the reasons that trouble accurate prediction of downstream oil product density monitoring include: the density of diesel and gasoline changes with changes in temperature and pressure, and complex operating conditions lead to density changes. It has a highly nonlinear relationship with hydrothermal power, that is, pressure and temperature data; hardware equipment generally faces zero point drift, which means that even under similar temperature and pressure conditions, upstream and downstream density meters may obtain different measurements for the same oil product. result. Therefore, if the subsequent oil monitoring density is directly predicted during modeling, the measurement errors inherent in hydrothermal conditions and different hardware equipment may lead to unreliable prediction results.

However, the oil products from the front and back have experienced similar hydrothermal processes during pipeline transportation, and the same equipment is used to measure their density information both upstream and downstream. Therefore, it is currently necessary to monitor the density difference at upstream and downstream stations of oil products. Difference in density monitoring from upstream and downstream oil product stations/> There must be a high degree of correlation between them. Use a data-driven model to capture the functional relationship between them, that is, monitor the density difference between the upstream and downstream stations of Qianxing Oil Products/> Difference in density monitoring from upstream and downstream oil product stations/> Model the dependence between the oil products and monitor the density difference at upstream and downstream oil product stations after correction/> , is a more reasonable modeling method.

In this process, the modal identification function is introduced by combining the GMR algorithm, so that the model can monitor the density difference of upstream and downstream oil stations under different modal conditions. Difference in density monitoring from upstream and downstream stations of Qianxing Oil Products/> Accurately model the functional relationship between them to ensure the accuracy of model prediction.

Equation (20) shows the theoretical formula for the change of oil density with temperature, where Characterizes the standard density of oil at 20 degrees Celsius, in kg/m3. In order to demonstrate the effectiveness and superiority of the modeling method proposed in this embodiment, theoretical formulas and two common machine learning algorithms are used, namely gradient boosted decision tree (Gradient Boosted DecisionTree) and artificial neural network algorithm (Artificial Neural network, ANN). ) and the prediction results obtained by directly using the GMR algorithm to directly predict the subsequent oil monitoring density. In addition, in order to evaluate the model performance as comprehensively as possible, 121 samples were divided into training sets and prediction sets, with the test set proportion gradually increasing from 0.3 to 0.6.

Tables 1 and 2 respectively show the subsequent oil monitoring density prediction accuracy of each model based on the root mean square error RMSE under the conditions of gasoline-diesel interface and diesel-gasoline interface. The error index calculation formula is as follows, where with/> Respectively represent the first/> The true values and fitted values of the samples,/> is the test sample size. The smaller the RMSE, the lower the overall error.

(20)

(twenty one)

Table 2 Prediction results of each model under the gasoline and diesel interface:

Table 3 Prediction results of each model under the diesel-gasoline interface:

It can be seen from Tables 2 and 3 that the downstream oil monitoring density prediction method proposed in this embodiment has obvious advantages, proving that the method proposed by the present invention is of great significance in improving the positioning of the mixed oil interface in product oil pipelines.

Embodiment 2

The purpose of this embodiment is to provide an oil product monitoring and density prediction system behind the mixed oil interface of a product oil pipeline, including:

The oil product monitoring and density prediction system behind the mixed oil interface of the refined oil pipeline is characterized by:

Acquisition module: obtain historical data of the mixed oil interface of the refined oil pipeline;

Gaussian module: Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the following oil products The difference in monitoring density between the upstream and downstream stations of the product is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained;

Training module: Based on the modal identification results, use the maximum expectation algorithm for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the modal;

Prediction module: Use the correction model of the monitoring density difference between the upstream and downstream stations of the downstream oil products to predict the mixed oil interface of the refined oil pipeline to be predicted, and combine it with the obtained monitoring density value of the upstream oil products in the refined oil pipeline to obtain the prediction Monitoring density value of downstream oil products at downstream stations of refined oil pipelines.

In this embodiment, the training module includes:

The first calculation module: calculates the posterior probability of the modal affiliation of the key input feature variables in each training sample;

The second calculation module: Based on the obtained posterior probability of the modal affiliation of the key feature variables in each training sample, calculate the variance, mean and covariance of the Gaussian distribution of each mode;

The third calculation module: Based on the calculated variance of the Gaussian distribution, establish the relationship between the key input feature variables and the output variables;

The fourth calculation module: Based on the calculated mean and covariance of the Gaussian distribution, establish the marginal probability density function of the key input feature variables.

In this embodiment, the prediction module includes:

Determination module: Use the temperature and pressure of the upstream and downstream stations of the refined oil pipeline to be predicted, and the front and rear oil product monitoring densities obtained by the upstream stations as the modal identification variables in the Gaussian mixture regression algorithm, and use the forward and backward oil product monitoring densities obtained by the upstream and downstream stations as the modal identification variables in the Gaussian mixture regression algorithm. The oil product monitoring density difference is used as the regressor variable in the Gaussian mixture regression algorithm;

Probability calculation module: calculates the posterior probability of the modal classification of the modal identification variable;

Difference prediction module: Based on the subsequent oil monitoring density difference correction model, the posterior probability of the modal classification of the modal identification variables, and the regression variables, the predicted value of the subsequent oil monitoring density difference is obtained.

In this embodiment, the prediction module also includes:

Addition module: Add the predicted density difference prediction value of the downstream oil product to the obtained monitoring density value of the downstream oil product upstream station of the refined oil pipeline to obtain the downstream oil product downstream station of the refined oil pipeline to be predicted Monitor the density value.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of the present invention. Those skilled in the art should understand that based on the technical solutions of the present invention, those skilled in the art do not need to perform creative work. Various modifications or variations that can be made are still within the protection scope of the present invention.

Claims (10)

1. The density prediction method of oil product monitoring behind the mixed interface of refined oil pipelines is characterized by: Obtain historical data of the mixed oil interface of product oil pipelines; Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream and downstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the upstream and downstream stations of the subsequent oil products. The station monitoring density difference is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained; According to the mode identification results, the maximum expectation algorithm is used for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the mode; The correction model of the density difference between upstream and downstream oil product monitoring stations is used to predict the mixed oil interface of the product oil pipeline to be predicted. Combined with the obtained monitoring density value of the oil product upstream station for the product oil pipeline, the product oil pipeline to be predicted is obtained. Monitoring density value of downstream oil products at downstream stations. 2. The method for predicting the density of oil product monitoring at the mixed interface of a refined oil pipeline as claimed in claim 1, wherein the hydraulic and thermal information of the upstream and downstream stations includes the temperature and pressure of the outlet of the upstream station and the inlet of the downstream station. Temperature and pressure; determine the monitoring density difference between the upstream and downstream stations of the forward oil product based on the density monitoring value of the upstream oil product pipeline and the density monitoring value of the forward oil product downstream station; based on the rear oil product pipeline density monitoring value The density monitoring value of the upstream station of the product and the density monitoring value of the downstream station of the subsequent oil product determine the monitoring density difference between the upstream and downstream stations of the subsequent oil product. 3. The method for monitoring and predicting the density of downstream oil products at the mixed interface of a refined oil pipeline as claimed in claim 1, characterized in that, based on the modal recognition results, the maximum expectation algorithm is used for training to obtain the downstream oil product density corresponding to the modality. Downstream station monitoring density difference correction model, specifically: Calculate the posterior probability of the modal affiliation of the key input feature variables in each training sample; Based on the obtained posterior probability of the modal affiliation of the key feature variables in each training sample, calculate the variance, mean and covariance of the Gaussian distribution of each modality; Based on the calculated variance of the Gaussian distribution, establish the relationship between the key input feature variables and the output variables; Based on the calculated mean and covariance of the Gaussian distribution, the marginal probability density function of the key input feature variables is established. 4. The method for predicting the density of oil product monitoring at the mixed interface of a refined oil pipeline as claimed in claim 1, characterized in that the mixed oil interface of the refined oil pipeline to be predicted is predicted using the density difference correction model of the oil product monitored at the rear. Specifically: The temperature and pressure of the upstream and downstream stations of the refined oil pipeline to be predicted, and the obtained monitoring density of the upstream and downstream oil products are used as the modal identification variables in the Gaussian mixture regression algorithm, and the obtained upstream and downstream station monitoring densities of the preceding oil products are The density difference is used as a regressor in the Gaussian mixture regression algorithm; Calculate the posterior probability of the modal affiliation of the modal identification variable; Based on the correction model of the density difference of downstream oil products monitoring, the posterior probability of the modal classification of the modal identification variables, and the regression variables, the predicted value of the monitoring density difference of downstream oil products downstream stations is obtained. 5. The method for predicting the density of oil product monitoring at the mixed interface of a refined oil pipeline as claimed in claim 4, characterized in that the density difference prediction value of the oil product monitoring at the rear line is compared with the obtained density difference of the oil product at the mixed interface of the refined oil pipeline. The monitoring density values of the upstream oil product stations are added up to obtain the monitoring density values of the downstream oil products downstream of the refined oil pipeline to be predicted. 6. The density prediction method for monitoring and predicting the density of downstream oil products at the oil-mixing interface of a refined oil pipeline as claimed in claim 4, characterized in that, according to the posterior probability and regression variables of the modal assignment of the modal identification variables, the downstream oil products are obtained The expectation of the density difference monitored by the downstream station is used as the predicted value of the density difference monitored by the downstream oil product downstream station. 7. The oil product monitoring and density prediction system behind the mixed oil interface of the refined oil pipeline is characterized by: Acquisition module: obtain historical data of the mixed oil interface of the refined oil pipeline; Gaussian module: Select the hydraulic and thermal information of the upstream and downstream stations of the refined oil pipeline in the historical data, the monitoring density of the upstream stations of the preceding oil products, and the monitoring density difference between the upstream and downstream stations of the preceding oil products as the key input feature variables, and select the following oil products The difference in monitoring density between the upstream and downstream stations of the product is used as the output variable, and the Gaussian mixture regression algorithm is used to perform multi-modal identification of key input feature variables and output variables, and the modal identification results in each mode are obtained; Training module: Based on the modal identification results, use the maximum expectation algorithm for training to obtain the density difference correction model for upstream and downstream oil product monitoring at the upstream and downstream stations corresponding to the modal; Prediction module: Use the correction model of the monitoring density difference between the upstream and downstream stations of the downstream oil products to predict the mixed oil interface of the refined oil pipeline to be predicted, and combine it with the obtained monitoring density value of the upstream oil products in the refined oil pipeline to obtain the prediction Monitoring density value of downstream oil products at downstream stations of refined oil pipelines. 8. The oil product monitoring and density prediction system behind the mixed interface of the refined oil pipeline according to claim 7, characterized in that the training module includes: The first calculation module: calculates the posterior probability of the modal affiliation of the key input feature variables in each training sample; The second calculation module: Based on the obtained posterior probability of the modal affiliation of the key feature variables in each training sample, calculate the variance, mean and covariance of the Gaussian distribution of each mode; The third calculation module: Based on the calculated variance of the Gaussian distribution, establish the relationship between the key input feature variables and the output variables; The fourth calculation module: Based on the calculated mean and covariance of the Gaussian distribution, establish the marginal probability density function of the key input feature variables. 9. The oil product monitoring and density prediction system behind the mixed interface of the refined oil pipeline according to claim 7, characterized in that the prediction module includes: Determination module: The temperature and pressure of the upstream and downstream stations of the refined oil pipeline to be predicted, and the front and rear oil product monitoring densities obtained by the upstream stations are used as modal identification variables in the Gaussian mixture regression algorithm, and the forward and backward oil product monitoring densities obtained by the upstream and downstream stations are used. The oil product monitoring density difference is used as the regressor variable in the Gaussian mixture regression algorithm; Probability calculation module: calculates the posterior probability of the modal classification of the modal identification variable; Difference prediction module: Based on the subsequent oil product monitoring density difference correction model, the posterior probability of modal identification variables and regression variables, the predicted value of the subsequent oil product monitoring density difference is obtained. 10. The density prediction system for monitoring the downstream oil product at the mixed interface of the refined oil pipeline as claimed in claim 7, characterized in that the prediction module further includes: an addition module: adding the density difference of the downstream oil product monitoring The predicted value is added to the obtained upstream station monitoring density value of oil products flowing behind the refined oil pipeline to obtain the monitoring density value of the downstream station monitoring density of oil products flowing behind the refined oil pipeline to be predicted.