Intelligent predictive regulation and control method and system for gas desulfurization system

Abstract

The invention provides an intelligent predictive regulation and control method and system for a gas desulfurization system, and belongs to the technical field of computers. The method comprises the following steps of: screening parameters from training data to form a plurality of structural equation models; and determining an optimal feature model from the plurality of structural equation models, and substituting model parameters of the optimal feature model into the prediction model. And training a prediction model by using the training data, and carrying out model optimization to determine an optimal prediction model. And predicting the hydrogen sulfide content of each stream of gas by using the optimal prediction model, and calculating the hydrogen sulfide content of the gas of the total pipe network by taking the flow data of each stream of gas as a weight. And if the hydrogen sulfide content of the gas in the main pipe network exceeds the standard, adjusting the gas according to an adjusting rule until the hydrogen sulfide content of the gas in the main pipe network does not exceed the standard so as to ensure that SO2 in the heating furnace reaches the standard and is discharged. The SO2 in the tail gas of the heating furnace is predicted in advance, process parameters of a gas desulfurization system are optimized and continuously adjusted in time, and stable and up-to-standard emission of SO2 is guaranteed from the source.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an intelligent prediction and control method and system for a gas desulfurization system. Background Art

[0002] A large amount of gas is generated during the production process of the refinery. These gases are usually at a very low pressure and cannot be directly used as fuel for the heating furnace. The traditional treatment method is to burn them in a flare, which not only causes a lot of energy waste, but also pollutes the environment. In recent years, the fuel supply for the refinery's heating furnace has mainly come from the gas produced as a by-product of each processing unit, which basically eliminates the random burning of gas in the flare and recycles the useful components in the gas. Low-pressure gas cannot be directly used in the gas pipeline network due to its low pressure, unstable flow and composition. Therefore, it is introduced into the gas cabinet through the flare discharge system for recovery, and then sent to the desulfurization device to remove hydrogen sulfide after being pressurized by a compressor, and finally enters the gas pipeline network for reuse. With the deterioration of crude oil and the increase of sulfur content, the sulfur content in gas also increases.

[0003] There are several methods for gas desulfurization, including wet desulfurization, dry desulfurization, direct oxidation and biological desulfurization. Chemical absorption, physical solvent absorption and chemical physical absorption in wet desulfurization are collectively referred to as renewable solvent methods, which are the most commonly used methods for desulfurization of petroleum, natural gas and refinery dry gas. In my country, the most widely used method is amine desulfurization. According to the number of hydrogen atoms replaced on the ammonia molecule, amines are mainly divided into primary amines, secondary amines and tertiary amines. The primary amine aqueous solution has the strongest alkalinity, followed by secondary amines and the weakest tertiary amines. Methyldiethanolamine (MDEA) solvent is currently widely used in industrial devices due to its advantages such as high concentration, high acid gas load, low corrosiveness, strong anti-degradation ability, high desulfurization selectivity and low energy consumption. However, the desulfurization solvents currently used all have the problem of poor removal effect on organic sulfur, and with the trend of crude oil deterioration, the sulfide and organic sulfur content in the raw gas may increase and exceed the maximum value of the process design.

[0004] As the fuel for the subsequent heating furnace, the high content of hydrogen sulfide in the gas will cause the sulfur content of the flue gas discharged by the heating furnace to exceed the standard. Many studies have focused on the optimization of the desulfurization device under steady-state conditions. For example, the amine liquid desulfurization model is established using the AspenHysys process simulation software to optimize the dosage of MDEA solution, ensure the H2S content in the desulfurized dry gas while saving the power consumption of the pump; another example is to use the Kent-Eisenberg model in AMSIM to simulate the dry gas desulfurization process, and analyze the many factors affecting the dry gas desulfurization effect in combination with the calculation results, optimize the operating temperature and operating pressure of the desulfurization tower, the dosage of lean amine liquid, the concentration of amine liquid, the H2S content in the lean amine liquid and the number of desulfurization tower plates, and adjust the acid gas content in the purified gas; another example is to use the AspenPlus software to simulate the process of the dry gas desulfurization system, and analyze the influence of factors such as gas-liquid ratio and temperature on the desulfurization effect. Another example is to use the radial basis artificial neural network based on the PSO algorithm to optimize the energy consumption under the premise of satisfying the H2S concentration constraint of the purified dry gas, and reduce the system energy consumption by 8.98% compared with the system energy consumption before optimization.

[0005] In view of the problems of frequent feed fluctuations and delayed optimization response in the heavy catalyst dry gas desulfurization process, a method for gas production data integration and intelligent prediction and control is urgently needed to provide support for gas desulfurization and help the back-end heating furnace exhaust gas meet emission standards. Summary of the invention

[0006] The purpose of the embodiments of the present invention is to provide an intelligent prediction and control method and system for a gas desulfurization system, so as to at least solve the above-mentioned problem of failing to predict and control the hydrogen sulfide content of the gas at the outlet of the heating furnace device.

[0007] In order to achieve the above-mentioned object, the first aspect of the present invention provides an intelligent prediction and control method for a gas desulfurization system, which comprises:

[0008] Parameters are screened from the preprocessed training data to form multiple structural equation models;

[0009] Determine the optimal feature model from multiple structural equation models, and substitute the model parameters of the optimal feature model into the prediction model as the modeling parameters of the prediction model, the prediction model includes a BP neural network model and a LSTM neural network model;

[0010] Use the preprocessed training data to train the BP neural network model and the LSTM neural network model respectively;

[0011] According to the preset model evaluation index, the optimal prediction model is determined from the trained BP neural network model and the trained LSTM neural network model;

[0012] The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network;

[0013] If the hydrogen sulfide content of the gas in the main pipeline network exceeds the standard, the gas with excessive hydrogen sulfide content shall be adjusted according to the regulation rules, and the hydrogen sulfide content of the gas in the main pipeline network shall be calculated again after adjustment. If the hydrogen sulfide content of the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content shall be adjusted again until the hydrogen sulfide content of the gas in the main pipeline network does not exceed the standard.

[0014] Optionally, the above-mentioned intelligent prediction and control method for the gas desulfurization system further includes:

[0015] According to the preset data update frequency, the hydrogen sulfide content of the main pipeline network gas is calculated and updated in real time.

[0016] Optionally, the above-mentioned intelligent prediction and control method for the gas desulfurization system further includes:

[0017] Based on the delivery time of gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network, it is predicted whether the delivered gas will cause the SO2 at the outlet of the heating furnace device to exceed the standard.

[0018] Optionally, the above-mentioned prediction of whether the transported gas will cause the outlet SO2 of the heating furnace device to exceed the standard based on the transport time of the gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network, further includes:

[0019] According to the length of the gas pipeline network and the gas transmission flow rate, the formula Calculate the delivery time for gas to reach each heating furnace device, where t is the delivery time, L is the length of the gas pipe network, and v is the gas delivery flow rate.

[0020] Optionally, the above parameter screening from the preprocessed training data includes:

[0021] Parameter screening is performed from the preprocessed training data through correlation analysis and least squares-based stepwise method.

[0022] Optionally, the above correlation analysis includes Pearson correlation analysis and Spearman correlation analysis;

[0023] Prior to parameter screening from the preprocessed training data by correlation analysis and least squares-based stepwise method, it also includes:

[0024] Perform normality analysis on the preprocessed training data;

[0025] If the preprocessed training data conforms to the normal distribution, Pearson correlation analysis is used;

[0026] If the preprocessed training data did not conform to a normal distribution, Spearman correlation analysis was used.

[0027] Optionally, the above-mentioned parameter screening is performed from the preprocessed training data to form multiple structural equation models, including:

[0028] Performing correlation analysis on the preprocessed training data to obtain correlation analysis results, and using a least squares-based stepwise method to select parameters from the preprocessed training data to obtain a least squares stepwise method analysis result;

[0029] According to the results of correlation analysis, the parameters that meet the requirements of correlation and collinearity are selected and included in the first structural equation model, and the included parameters are classified as latent variables according to pressure, flow rate and temperature;

[0030] According to the least squares stepwise analysis results, the model parameters that meet the requirements of multivariate and goodness of fit are selected and included in the second structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and hydraulic control;

[0031] Combining the results of correlation analysis and least squares stepwise analysis, common parameters were screened and included in the third structural equation model, and the included parameters were classified as latent variables according to pressure, flow, and hydraulic control.

[0032] Optionally, the above-mentioned intelligent prediction and control method for the gas desulfurization system further includes:

[0033] Obtain training data collected from the DCS of the gas desulfurization system.

[0034] Optionally, the above-mentioned intelligent prediction and control method for the gas desulfurization system further includes:

[0035] Preprocess the training data.

[0036] Optionally, the above preprocessing of the training data includes:

[0037] Perform outlier screening and difference processing on the training data, and delete null value data;

[0038] Resample the training data.

[0039] Optionally, the training data collected from the DCS of the gas desulfurization system includes:

[0040] According to the correlation between the SO2 data at the outlet of the heating furnace device and the control parameters of the hydrogen sulfide content in the gas, multiple influencing indicators for controlling the hydrogen sulfide content in the gas are determined;

[0041] Obtain training data collected from the DCS based on multiple impact indicators.

[0042] Optionally, the above-mentioned determining the optimal characteristic model from multiple structural equation models includes:

[0043] The chi-square freedom ratio, goodness-of-fit index, and root mean square error of approximation of each structural equation model were compared to determine the optimal characteristic model.

[0044] Optionally, the above-mentioned using the preprocessed training data to respectively train the BP neural network model and the LSTM neural network model includes:

[0045] Divide the preprocessed training data into a training set and a test set according to a preset ratio;

[0046] The training set is used to train the BP neural network model and the LSTM neural network model respectively, and the test set is used to verify the BP neural network model and the LSTM neural network model.

[0047] Optionally, the above preset model evaluation indicators include root mean square error RMSE and determination coefficient R 2 .

[0048] Optionally, the above-mentioned determining the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index includes:

[0049] The root mean square error RMSE and the coefficient of determination R 2 The neural network model with a higher value is taken as the optimal prediction model.

[0050] Optionally, the above optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network. The calculation formula is as follows:

[0051] CS=Q1 / (Q1+Q2+…+Qn)×CS1+Q2 / (Q1+Q2+…+Qn)×CS2+…+Qn / (Q1+Q2+…+Qn)×CSn, wherein CS is the hydrogen sulfide content of the gas in the main pipeline network, CS1, CS2, ...CSn are the hydrogen sulfide contents of each gas stream, Q1, Q2, ...Qn are the flow rates of each gas stream, and Q1 / (Q1+Q2+…+Qn), ...Qn / (Q1+Q2+…+Qn)×CSn are the weight coefficients of each gas stream respectively.

[0052] Optionally, the above-mentioned adjustment rules are used to adjust the gas with excessive hydrogen sulfide content, including:

[0053] Reduce the gas inlet flow rate of the gas desulfurization system and increase the amine liquid circulation volume of the gas desulfurization system in steps of 1%.

[0054] The second aspect of the present invention provides an intelligent prediction and control system for a gas desulfurization system, which comprises:

[0055] A parameter screening module is used to screen parameters from the preprocessed training data to form multiple structural equation models;

[0056] An optimal feature model determination module is used to determine the optimal feature model from multiple structural equation models, and substitute the model parameters of the optimal feature model into the prediction model as modeling parameters incorporated into the prediction model, and the prediction model includes a BP neural network model and an LSTM neural network model;

[0057] The model training module is used to train the BP neural network model and the LSTM neural network model respectively using the preprocessed training data;

[0058] The optimal prediction model determination module is used to determine the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index;

[0059] The hydrogen sulfide content prediction module is used to predict the hydrogen sulfide content of each gas stream using the optimal prediction model, and use the flow data of each gas stream as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network;

[0060] The gas adjustment module is used to adjust the gas with excessive hydrogen sulfide content in the main pipeline network according to the adjustment rules if the hydrogen sulfide content in the gas exceeds the standard. After the adjustment, the hydrogen sulfide content in the gas in the main pipeline network is calculated again. If the hydrogen sulfide content in the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content is adjusted again until the hydrogen sulfide content in the gas in the main pipeline network does not exceed the standard.

[0061] In a third aspect of the present invention, a machine-readable storage medium is provided, on which instructions are stored. When the instructions are executed by a processor, the processor is configured to execute the intelligent prediction and control method for a gas desulfurization system.

[0062] In a fourth aspect of the present invention, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the intelligent prediction and control method for the gas desulfurization system is implemented.

[0063] Through the above technical scheme, a method and system for intelligent prediction and control of gas desulfurization system obtains training data collected from DCS and preprocesses the training data to avoid the influence of null values ​​in the training data, while keeping the frequency of data acquisition consistent. Through correlation analysis and the step-by-step method based on least squares, parameters are screened from the preprocessed training data to form multiple structural equation models, and the model parameters of the optimal characteristic model are selected from the multiple structural equation models and substituted into the prediction model. The preprocessed training data is used to train the BP neural network model and the LSTM neural network model respectively, and the model is optimized from the trained BP neural network model and the trained LSTM neural network model to obtain the optimal prediction model. The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network, thereby predicting whether the SO2 of the heating furnace tail gas meets the standard. According to the regulation rules, the gas flow rate and the amine liquid circulation volume are adjusted to adjust the operating parameters of the gas desulfurization system in advance, so as to ensure that the SO2 emissions of the rear-end heating furnace exhaust gas meet the emission standards. This method and system realize the advance prediction of SO2 in the heating furnace exhaust gas, and timely provide a continuous adjustment plan for the optimization of the process parameters of the gas desulfurization system, so as to ensure the stable and standard emission of SO2 from the source.

[0064] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0066] Figure 1 It is a flow chart of a method for intelligent prediction and control of a gas desulfurization system provided by one embodiment of the present invention;

[0067] Figure 2 It is a schematic diagram of the process of establishing a heating furnace tail gas SO2 compliance control technology based on the prediction and optimization of the hydrogen sulfide content in the front-end gas pipe network provided by an embodiment of the present invention;

[0068] Figure 3 is a schematic diagram of a first structural equation model provided by an embodiment of the present invention;

[0069] Figure 4 is a schematic diagram of a second structural equation model provided by one embodiment of the present invention;

[0070] Figure 5 is a schematic diagram of a third structural equation model provided by an embodiment of the present invention;

[0071] Figure 6 It is a result diagram of a gas pipe network hydrogen sulfide content prediction model provided by an embodiment of the present invention;

[0072] Figure 7 It is another result diagram of a gas pipe network hydrogen sulfide content prediction model provided by an embodiment of the present invention;

[0073] Figure 8 It is a structural block diagram of an intelligent prediction and control system for a gas desulfurization system provided by an embodiment of the present invention;

[0074] Fig. 9 is a schematic diagram of a Spearman correlation test provided by one embodiment of the present invention;

[0075] Fig.10 It is a schematic diagram of the structure of an electronic device provided by a preferred embodiment of the present invention.

[0076] Description of Reference Numerals

[0077] 10 - electronic device, 100 - processor, 101 - memory, 102 - computer program. DETAILED DESCRIPTION

[0078] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0079] Example 1

[0080] Please refer to Figure 1 and Figure 2 , Figure 1 It is a flow chart of a method for intelligent prediction and control of a gas desulfurization system provided by one embodiment of the present invention. Figure 2 The present invention provides a method for intelligent prediction and control of gas desulfurization system, which includes:

[0081] S110: screening parameters from the preprocessed training data to form multiple structural equation models;

[0082] Optionally, the above-mentioned parameter screening from the preprocessed training data includes: screening the parameters from the preprocessed training data through correlation analysis and a least squares-based stepwise method.

[0083] Specifically, correlation analysis and the least squares-based stepwise method are used to select features from the preprocessed training data, and multiple structural equation models are formed using the selected feature parameters. The purpose of establishing multiple structural equation models based on multiple parameter optimization methods is achieved.

[0084] Among them, the structural equation model is a general and effective linear modeling method, which includes a variety of statistical analysis methods and means such as factor analysis, multiple regression and path analysis. The structural equation model overcomes the defect that the measurement variables in the simultaneous equation model cannot have errors, and allows errors to exist in both the independent variables and the dependent variables. There are two types of variables in the structural equation model, one is the latent variable, that is, the variable that cannot be directly measured; the other is the measurement variable. There are two types of models in the structural equation model: measurement model and structural model. The measurement model calculates the relationship between the measured variables and the latent variables, and the structural model calculates the relationship between the latent variables. Its expression equation is:

[0085] Measurement equation

[0086] X=Λ_xξ+δ

[0087] Y=Λ_yη+ε

[0088] Structural equation

[0089] η=Bη+Γξ+ζ

[0090] The measurement equation is the equation of the relationship between the observed variables X, Y and the latent variables η, ξ. The structural equation is the equation of the relationship between the latent variables. X represents the exogenous variable that is not affected by other variables, and Y represents the endogenous variable that is affected by any variable in the model. Λ_x represents the relationship between the exogenous indicator and the exogenous latent variable, ξ is the exogenous latent variable, and δ is the measurement error; Λ_y represents the relationship between the endogenous indicator and the endogenous latent variable, η is the endogenous latent variable, and ε is the measurement error; B is the relationship between the endogenous latent variables, Γ is the impact of the exogenous latent variable on the endogenous latent variable, and ζ is the residual of the equation structure.

[0091] The intelligent prediction and control method for the gas desulfurization system further comprises: obtaining training data collected from the DCS of the gas desulfurization system.

[0092] In some implementations of this embodiment, the above-mentioned acquisition of training data collected from the DCS of the gas desulfurization system includes: determining multiple influencing indicators for controlling the hydrogen sulfide content of the gas according to the correlation between the SO2 data at the outlet of the heating furnace device and the control parameters of the hydrogen sulfide content of the gas. The training data collected from the DCS is acquired according to the multiple influencing indicators.

[0093] Among them, the above-mentioned influencing indicators may include gas flow rate, desulfurization tower inlet pressure, desulfurization tower outlet pressure, separator tank low pressure, high pressure and other equipment gas pressure, desulfurization tower inlet amine liquid pressure, amine liquid tank level, amine liquid into the desulfurization tower flow, desulfurization tower bottom liquid control, desulfurization tower outlet separator tank level.

[0094] Specifically, based on the correlation between the SO2 data at the outlet of the heating furnace device and the control parameters of the hydrogen sulfide content in the gas, the influencing indicators such as the amine liquid flow rate, desulfurization tower pressure, and desulfurization tower temperature of the gas desulfurization system that control the hydrogen sulfide content in the gas are clarified, so as to obtain the training data collected from the DCS based on the influencing indicators.

[0095] The above-mentioned intelligent prediction and control method for the gas desulfurization system also includes: preprocessing the training data.

[0096] In some implementations of this embodiment, the preprocessing of the training data includes: screening outliers and performing difference processing on the training data, deleting null value data, and resampling the training data.

[0097] Specifically, according to the influencing indicators, the training data obtained include process parameters such as amine liquid flow rate, desulfurization tower pressure, and desulfurization tower temperature of the gas desulfurization system. In order to avoid the influence of null values ​​in the training data, the training data is screened for abnormal values ​​and difference processing is performed to delete the null value data. And because the initial sampling data of the process parameters of the gas desulfurization system have inconsistent frequencies, in order to ensure the smooth progress of subsequent data modeling, data resampling is performed to keep the data acquisition frequency consistent.

[0098] S120: determining an optimal feature model from multiple structural equation models, and substituting model parameters of the optimal feature model into the prediction model as modeling parameters incorporated into the prediction model, the prediction model including a BP neural network model and a LSTM neural network model;

[0099] Among them, the prediction model includes BP neural network and long short-term memory neural network (Long Short-Term Memory, LSTM). BP neural network is a multi-layer feedforward neural network, consisting of an input layer, a hidden layer, and an output layer. A 3-layer BP neural network model can highly fit any function, can learn the nonlinear relationship between gas desulfurization system parameters and hydrogen sulfide content, and predict the hydrogen sulfide content of gas when the error is minimized. The LSTM model is a neural network improved on the basis of the recurrent neural network, which makes up for the defect that RNN cannot process long sequence data. The LSTM neural network controls the transmission of information by adding memory gates, learning gates, and forgetting gates.

[0100] In some implementations of this embodiment, the above-mentioned determining the optimal characteristic model from multiple structural equation models includes: comparing the chi-square freedom ratio, goodness of fit index and root mean square error of each structural equation model to determine the optimal characteristic model.

[0101] Specifically, all the established structural equation models were compared, and the chi-square degree of freedom ratio (Chi-square / df), goodness of fit index (GFI), and root mean square error of approximation (RMSEA) of each structural equation model were compared. The comparison criteria of each structural equation model were Chi-square / df and RMSEA values, the smaller the better, and the GFI value, the larger the better, so as to determine the optimal characteristic model, and select the model parameters of the optimal characteristic model as the modeling parameters included in the prediction model and substitute them into the prediction model. In this way, the path analysis of the structural equation model was used to optimize the model parameters, reduce the collinearity of the parameters, and optimize the modeling parameters from a large number of parameters. The purpose of combining the optimal characteristic model with the prediction model was achieved, thereby realizing the screening and optimization of the modeling parameters of the prediction model and improving the accuracy of the modeling.

[0102] S130: using the preprocessed training data to train the BP neural network model and the LSTM neural network model respectively;

[0103] Specifically, for the BP neural network model and the LSTM neural network model, the input is based on the characteristic latent variables of the preferred structural equation model, and the output is the hydrogen sulfide content of the gas. The BP neural network model and the LSTM neural network model are trained respectively according to the preprocessed training data, and the predicted value of the hydrogen sulfide content of the gas is calculated based on the trained neural network model using the data collected in the DCS, and verified using the preprocessed training data. Thus, a trained BP neural network model and a trained LSTM neural network model are obtained.

[0104] S140: Determine the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index;

[0105] Among them, the above preset model evaluation indicators include root mean square error RMSE and determination coefficient R 2 .

[0106] In some implementations of this embodiment, the above-mentioned determining the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index includes: calculating the root mean square error RMSE and the determination coefficient R 2 The neural network model with a higher value is taken as the optimal prediction model.

[0107] Specifically, according to the root mean square error RMSE and determination coefficient R of the two neural network models 2 Numerical value, model screening is performed in the trained BP neural network model and the trained LSTM neural network model, R 2 The higher the RMSE value, the better the model fit.

[0108] S150: using the optimal prediction model to predict the hydrogen sulfide content of each gas stream, and using the flow data of each gas stream as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network;

[0109] Exemplarily, the above-mentioned gas streams may include catalytic cracking stripped gas, S-Zorb stripped gas, hydrogenation unit stripped gas, wax oil hydrogenation unit stripped gas and low-pressure gas stripped gas.

[0110] Furthermore, the above-mentioned optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow rate data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the total pipeline network according to the following formula: CS=Q1 / (Q1+Q2+…+Qn)×CS1+Q2 / (Q1+Q2+…+Qn)×CS2+…+Qn / (Q1+Q2+…+Qn)×CSn, wherein CS is the hydrogen sulfide content of the gas in the total pipeline network, CS1, CS2, ...CSn are the hydrogen sulfide contents of each gas stream, Q1, Q2, ...Qn are the flow rates of each gas stream, and Q1 / (Q1+Q2+…+Qn), ...Qn / (Q1+Q2+…+Qn)×CSn are the weight coefficients of each gas stream, respectively.

[0111] Specifically, the hydrogen sulfide content of each gas stream is first predicted by the optimal prediction model, and then the flow data of the gas network management system is obtained according to the predicted source of each gas stream, and the flow data of each gas stream is used as the weight, and the hydrogen sulfide content of the gas stream in the total network is calculated according to the formula CS=Q1 / (Q1+Q2+…+Qn)×CS1+Q2 / (Q1+Q2+…+Qn)×CS2+…+Qn / (Q1+Q2+…+Qn)×CSn. Since Q1 / (Q1+Q2+…+Qn), ...Qn / (Q1+Q2+…+Qn)×CSn in the formula are the weight coefficients of each gas stream, the weight can be changed according to the change of flow rate through the above formula.

[0112] Optionally, the intelligent prediction and control method for the gas desulfurization system further includes: updating and calculating the hydrogen sulfide content of the gas in the main pipeline network in real time according to a preset data update frequency.

[0113] In this embodiment, the preset data update frequency is once every 3 minutes. According to the DCS data update frequency, the optimal prediction model runs once every 3 minutes, and the hydrogen sulfide content of each gas stream predicted by the optimal prediction model will also be updated once every 3 minutes, and the hydrogen sulfide content of the gas in the main pipeline network will also be updated along with the predicted hydrogen sulfide content of each gas stream.

[0114] In some implementations of this embodiment, the above-mentioned intelligent prediction and control method of the gas desulfurization system also includes: predicting whether the transported gas will cause the SO2 at the outlet of the heating furnace device to exceed the standard based on the transportation time of the gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network.

[0115] The above-mentioned transportation time of the gas to each heating furnace device refers to the transportation time of the gas from the gas cabinet or the desulfurization device to each heating furnace device.

[0116] Specifically, if the delivery time is t, then based on the delivery time t combined with the real-time updated hydrogen sulfide content of the total pipeline network gas, the hydrogen sulfide content of the total pipeline network gas after t time can be predicted, that is, whether the gas delivered from the gas cabinet or desulfurization device after t minutes will cause the SO2 at the outlet of the heating furnace device to exceed the standard.

[0117] S160: If the hydrogen sulfide content of the gas in the main pipeline network exceeds the standard, the gas with excessive hydrogen sulfide content will be adjusted according to the regulation rules. After the adjustment, the hydrogen sulfide content of the gas in the main pipeline network will be calculated again. If the hydrogen sulfide content of the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content will be adjusted again until the hydrogen sulfide content of the gas in the main pipeline network does not exceed the standard.

[0118] In some implementations of this embodiment, the above-mentioned adjustment rules are used to adjust the gas with excessive hydrogen sulfide content, including: reducing the gas inlet flow rate of the gas desulfurization system and increasing the amine liquid circulation volume of the gas desulfurization system in steps of 1%.

[0119] Specifically, when the hydrogen sulfide content of the gas in the main pipeline network exceeds the standard, it can be predicted that the SO2 at the outlet of the heating furnace device exceeds the standard. According to the comparison between the predicted hydrogen sulfide content of the gas in the main pipeline network and the hydrogen sulfide content control index, the gas with the excessive hydrogen sulfide content of the gas is adjusted, the gas inlet flow of the gas desulfurization system is reduced, and the amine liquid circulation volume of the gas desulfurization system is adjusted according to the step size of 1%. After the adjustment, S150 is executed again to calculate the adjusted hydrogen sulfide content of the gas in the main pipeline network. If the hydrogen sulfide content of the gas in the main pipeline network still exceeds the standard, S160 is repeatedly executed until the hydrogen sulfide content of the gas in the main pipeline network does not exceed the standard. In this way, the hydrogen sulfide content of the gas after desulfurization of the gas desulfurization system is predicted in advance, the hydrogen sulfide content of the gas in the main pipeline network is perceived by using a formula with a variable weight, and the gas inlet flow and the amine liquid circulation volume are gradually adjusted based on the adjustment rules to ensure that the SO2 of the tail gas of the rear heating furnace meets the emission standards.

[0120] In the above implementation process, the method first obtains the training data collected from the DCS and preprocesses the training data to avoid the influence of null values ​​in the training data and keep the frequency of data acquisition consistent. Then, through correlation analysis and the least squares-based step-by-step method, parameters are screened from the preprocessed training data to form multiple structural equation models, and the model parameters of the optimal feature model are selected from the multiple structural equation models and substituted into the prediction model. Then, the BP neural network model and the LSTM neural network model are trained respectively using the preprocessed training data, and the model is optimized from the trained BP neural network model and the trained LSTM neural network model to obtain the optimal prediction model. The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network, thereby predicting whether the SO2 exhaust gas of the heating furnace meets the standard. According to the regulation rules, the gas flow rate and the amine liquid circulation volume are adjusted to adjust the operating parameters of the gas desulfurization system in advance, so as to ensure that the SO2 emissions of the rear-end heating furnace exhaust gas meet the emission standards. This method realizes the early prediction of SO2 in the heating furnace exhaust gas, and timely provides a continuous adjustment plan for the optimization of the process parameters of the gas desulfurization system, ensuring the stable and standard emission of SO2 from the source.

[0121] Optionally, the above correlation analysis includes Pearson correlation analysis and Spearman correlation analysis;

[0122] Prior to parameter screening from the preprocessed training data by correlation analysis and least squares based walking method, the method further includes:

[0123] Perform normality analysis on the preprocessed training data;

[0124] If the preprocessed training data conforms to the normal distribution, Pearson correlation analysis is used;

[0125] If the preprocessed training data did not conform to a normal distribution, Spearman correlation analysis was used.

[0126] Specifically, the preprocessed training data is subjected to normality analysis before correlation analysis. If the data conform to normal distribution, Pearson correlation analysis is used. If the data do not conform to normal distribution, Spearman correlation analysis is used.

[0127] Optionally, the above-mentioned parameter screening is performed from the preprocessed training data to form multiple structural equation models, including:

[0128] Performing correlation analysis on the preprocessed training data to obtain correlation analysis results, and using a least squares-based stepwise method to select parameters from the preprocessed training data to obtain a least squares stepwise method analysis result;

[0129] Among them, the specific process of using the least squares-based stepwise method to screen parameters from the preprocessed training data is as follows: In order to avoid the influence of multicollinearity between indicators, the stepwise method first selects the variables that do not meet the significance test in the sample data that affect the hydrogen sulfide content of gas, and eliminates them, and then substitutes the next variable into the analysis model, and uses the least squares method to obtain a best fitting line. For the observed data, the stepwise method calculates the best fitting line by minimizing the sum of squares of vertical deviations from each data point to the line. Stepwise regression fits the model by adding covariates based on specified standards at the same time. It not only considers the test when the variable enters the regression model to select qualified variables, but also considers the collinearity problem between variables after the variables are introduced into the model. The model involves 16 influencing parameters, and the n of the inherited model is 16. The specific steps are as follows:

[0130] (1) Establish a univariate regression model of the dependent variable and the regression independent variable, obtain the independent variable Xi (i = 1, 2, ..., n), and calculate the regression coefficient F statistics F1(1), F2(1), ..., Fn(1), select the largest statistical value Fi1(1), and use the proposed critical value F for significance calculation. If Fi1

[0131] (1)>F, then the set of independent variables Ii is included in the model.

[0132] (2) Construct a binary regression model of the dependent variable Y and n-1 independent variable subsets {Xi1, X1}, …, {Xi1, Xi1-1}, {Xi1, Xi1+1}, …, {Xi1, Xn} in the regression model. Let Fk(2)k≠I1 be the statistical value of the corresponding regression coefficient F test, and let the maximum value be Fi2(2). Calculate the significance again and get the critical value as F2. If Fi2(2)>F2, include variable Xi2 in the regression model. If not, there is no need to introduce the next independent variable and end this round of calculation.

[0133] (3) Construct regression models with different dependent and independent variables and repeat the above steps. When all variables are introduced into the model and a new significance test is performed, the stepwise regression analysis ends.

[0134] According to the results of correlation analysis, the parameters that meet the requirements of correlation and collinearity are selected and included in the first structural equation model, and the included parameters are classified as latent variables according to pressure, flow rate and temperature;

[0135] According to the least squares stepwise analysis results, the model parameters that meet the requirements of multivariate and goodness of fit are selected and included in the second structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and hydraulic control;

[0136] Combining the results of correlation analysis and least squares stepwise analysis, common parameters were screened and included in the third structural equation model, and the included parameters were classified as latent variables according to pressure, flow, and hydraulic control.

[0137] Specifically, according to the results of the correlation analysis, the parameters with high correlation and small collinearity are screened and included in the structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and temperature, and all the screened dependent variables, that is, the indicator data with high correlation with hydrogen sulfide content screened out through the correlation analysis, are covered to establish the first structural equation model. According to the results of the least squares stepwise analysis, the model parameters with multiple variables and high fitting degree are screened and included in the structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and hydraulic control, and all the dependent variables, that is, the indicator data screened out through the least squares stepwise method, are covered to establish the second structural equation model. Combining the results of the correlation analysis and the results of the least squares stepwise analysis, the common parameters are screened and included in the structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and hydraulic control, and all the dependent variables are covered to establish the third structural equation model.

[0138] Optionally, the above-mentioned using the preprocessed training data to respectively train the BP neural network model and the LSTM neural network model includes:

[0139] Divide the preprocessed training data into a training set and a test set according to a preset ratio;

[0140] The training set is used to train the BP neural network model and the LSTM neural network model respectively, and the test set is used to verify the BP neural network model and the LSTM neural network model.

[0141] Exemplarily, the preset ratio may be 8:2, and the training set and the test set are divided according to the ratio of 8:2.

[0142] Optionally, before predicting whether the transported gas will cause the outlet SO2 of the heating furnace device to exceed the standard based on the transport time of the gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network, the method further includes:

[0143] According to the length of the gas pipeline network and the gas transmission flow rate, the formula Calculate the delivery time for gas to reach each heating furnace device, where t is the delivery time, L is the length of the gas pipe network, and v is the gas delivery flow rate.

[0144] Specifically, t is the transportation time of gas from the gas cabinet or desulfurization device to each heating furnace device, and L is the length of the gas pipeline network from the gas cabinet or desulfurization device to each heating furnace device.

[0145] Example 2

[0146] Figure 2 This is a schematic diagram of the process of establishing a heating furnace tail gas SO2 compliance control technology based on the prediction and optimization of the hydrogen sulfide content in the front-end gas network provided by an embodiment of the present invention. Figure 2 As shown, a gas desulfurization system intelligent prediction and control method can realize the intelligent prediction of the hydrogen sulfide content of the gas at the outlet of the cracking furnace gas desulfurization system. The specific steps are as follows:

[0147] Step S101: Acquire training data collected from a distributed control system (DCS) of a gas desulfurization system.

[0148] Step S102: The initial sampling data of the process parameters of the desulfurization device has a frequency inconsistency problem. In order to ensure the smooth data modeling, the data needs to be downsampled. The original data sampling frequency is shown in Table 1:

[0149] Table 1 Raw data sampling frequency

[0150]

[0151]

[0152] Since the 3-minute interval meets all data sampling intervals, the downsampled data interval is 3 minutes. There are a lot of null values ​​in the flare gas desulfurization tower lean liquid inlet flow control, flare gas desulfurization tower pressure, and flare gas desulfurization tower lower temperature data, which will affect the results during data analysis, and the missing data values ​​need to be removed or supplemented.

[0153] S103: Since the data do not conform to the normal distribution, Spearman correlation is used to perform bivariate correlation test. Fig. 9 is a schematic diagram of a Spearman correlation test provided by one embodiment of the present invention, such as Fig. 9As shown in the figure, the parameters with the greatest correlation with hydrogen sulfide content are: FIC1001, FIQ1001, PIA1723, FIC136, LIC1723, PI1102, PI1103, LIC1135, PIC134, PIC121, PIC138, LI13. Among them, there is a significant positive correlation with FIQ1001, LIC1723, and LI137, and a negative correlation with other parameters. It can be seen that MDEA reacts relatively quickly to H2S, which is an instantaneous reaction controlled by gas film. At lower temperature and higher pressure, the reaction proceeds to the right, and vice versa. From the correlation analysis, it can be seen that a larger gas pressure and a higher hydrogen sulfide outlet temperature are conducive to reducing the hydrogen sulfide content of the gas. At the same time, based on the Spearman correlation analysis, the collinearity between indicators can also be seen. It can be seen that PI1102 and PIA1723 have high collinearity (0.99); PI1103 and PIA1723 have high collinearity (0.99); PIC134 has high collinearity with PIA1723, PI1102, and PI1103, which are 0.95, 0.98, and 0.99, respectively.

[0154] S104: 16 multiple linear models were fitted using the stepwise method. The summary of the stepwise method model screening is shown in Table 2. By comparing the adjusted R 2 (excluding the influence of independent variables), we can see that the R of the last model is 2 The value is the highest. Then refer to the significance test results of the model and regression coefficient to determine which model is the best. It can be seen that the fitting R 2 Not much difference.

[0155] Table 2 Summary of stepwise model screening

[0156]

[0157] The significance of the 16 models is less than 0.01, reaching the significance level, indicating that all 16 models are meaningful. At the same time, the partial regression coefficient, partial correlation coefficient and collinearity statistical tolerance of the independent variables excluded in each model are analyzed to help determine whether there is a collinear relationship between the independent variables. It can be seen from the degree of fit that model 11 has achieved the best goodness of fit, which is R 0.01 higher than that of model 16. 2The difference is less than 0.001. To avoid collinearity, the parameters in model 11 can be selected. Model 11 includes VAR00006, VAR00002, VAR00018, VAR00015, VAR00014, VAR00019, VAR00017, VAR00016, VAR00010, VAR00013, VAR00020, respectively FIC136, FIC1001, LI137, PIC121, FI04, FI40002, PIC138, PIC137, TI1104, FI128, PIC40002. Among them, FIC136, FIC1001, LI137, PIC121, FI04 and TI1104 are the ones with higher firmness.

[0158] S105: The study used Spearman correlation and least squares-based stepwise method parameter screening to form three structural equation models. Figure 3 is a schematic diagram of a first structural equation model provided by an embodiment of the present invention, such as Figure 3 As shown, according to the results of the Spearman correlation coefficient analysis, the parameters with high correlation and small collinearity were screened and included in the first structural equation model, and the included parameters were classified as latent variables according to pressure and flow, and temperature and dependent variables were covered. Figure 4 is a schematic diagram of a second structural equation model provided by an embodiment of the present invention, such as Figure 4 As shown, according to the least squares stepwise analysis results, the model parameters with multiple variables and high fitting degree were screened and included in the second structural equation model. The included parameters were classified as latent variables according to pressure, flow, and hydraulic control, and temperature and dependent variables were covered. Figure 5 is a schematic diagram of a third structural equation model provided by an embodiment of the present invention, such as Figure 5 As shown, combined with the analysis results of Spearman correlation coefficient and least squares stepwise method, the common parameters were screened and included in the third structural equation model, and the included parameters were classified as latent variables according to pressure, flow, and hydraulic control, and temperature and dependent variables were included; the evaluation index results of the structural equation model are shown in Table 3:

[0159] Table 3 Comparison of evaluation indicators of structural equation model

[0160] index Model 1 Model 2 Model 3 Chi-square / df 3.9557 3.6643 2.9500 GFI 0.810 0.773 0.925 RMSEA 0.197 0.190 0.170

[0161] In Table 3, Model 1 is the first structural equation model, Model 2 is the second structural equation model, and Model 3 is the third structural equation model. It can be seen that the third structural equation model is the optimal model.

[0162] S106: Substitute the model parameters of the third structural equation model into the neural network model for repeated training, merge the sequence of predicted hydrogen sulfide content of gas at the outlet of the desulfurization device with the sequence of selected influencing factors, divide the training set and the test set into a ratio of 8:2, and input the training set and the test set into the BP neural network and LSTM neural network model training. Among them, the model parameters of the third structural equation model are shown in Table 4:

[0163] Table 4 Structural equation model parameters

[0164]

[0165]

[0166] S107: The model parameters are changed multiple times, and multiple groups (e.g., 13 groups) of model training are performed to obtain the optimal prediction model of hydrogen sulfide content in gas based on BP neural network, with the number of neurons (hidden layers) being 6, the number of iterations being 700, and the learning rate being 0.2. The RMSE is 2.98, and the prediction accuracy (RMSE / value range) is: 83.30%.

[0167] S108: LSTM model training parameters include: epoch, batchsize, learningrate, step size, number of hidden layers, random loss rate and cross-validation fold. After training, the hyperparameter tuning results are {'batch_size':8, 'epochs':3, 'optimizer':'adam'}, the model time step is 30, the input is a 13-dimensional vector, the number of hidden layers is 2, the random loss rate is 0.2, and the cross-validation fold is 2. The model results show that the RMSE is 1.26 and the prediction accuracy (RMSE / value range) is: 89.91%.

[0168] S109: Use the optimal prediction model to predict the hydrogen sulfide content of each gas stream. Figure 6 FIG. 1 is a result diagram of a gas pipeline network hydrogen sulfide content prediction model provided by an embodiment of the present invention. Figure 7 Shown is another result diagram of the prediction model for hydrogen sulfide content in a gas network provided by an embodiment of the present invention. Based on the gas flow rate of the gas cabinet, the gas flow rate of the constant pressure reduction device, the gas flow rate of the catalytic device, the gas flow rate of the hydrogen production device, the gas flow rate of the Szorb device, etc., the flow rate is used as a weight, and the hydrogen sulfide content of the gas in the total network is calculated based on the predicted hydrogen sulfide content of each gas stream from the main source. The weight is changed according to the change in flow rate, and the data is updated every 3 minutes. The transportation time of the gas from the gas cabinet to the heating furnace device is calculated as t based on the length and flow rate of the pipeline network. Through this prediction model, it can be predicted whether the SO2 at the outlet of the heating furnace device exceeds the standard after t minutes.

[0169] S110: When it is predicted that SO2 exceeds the standard, the gas exceeding the standard is adjusted based on the comparison between the predicted value and the hydrogen sulfide content control index, the gas inlet flow of the desulfurization system is reduced, and the amine liquid circulation volume of the desulfurization system is increased. The step size is 1%. After adjustment, the hydrogen sulfide content of the gas in the total pipeline network is predicted until the weighted hydrogen sulfide content of the gas in the total pipeline network meets the standard.

[0170] Example 3

[0171] Figure 8 FIG. 1 is a block diagram of an intelligent prediction and control system for a gas desulfurization system provided by an embodiment of the present invention. Figure 8 As shown, an embodiment of the present invention provides an intelligent prediction and control system for a gas desulfurization system, which includes:

[0172] A parameter screening module is used to screen parameters from the preprocessed training data to form multiple structural equation models;

[0173] An optimal feature model determination module is used to determine the optimal feature model from multiple structural equation models, and substitute the model parameters of the optimal feature model into the prediction model as modeling parameters incorporated into the prediction model, and the prediction model includes a BP neural network model and an LSTM neural network model;

[0174] The model training module is used to train the BP neural network model and the LSTM neural network model respectively using the preprocessed training data;

[0175] The optimal prediction model determination module is used to determine the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index;

[0176] The hydrogen sulfide content prediction module is used to predict the hydrogen sulfide content of each gas stream using the optimal prediction model, and use the flow data of each gas stream as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network;

[0177] The gas adjustment module is used to adjust the gas with excessive hydrogen sulfide content in the main pipeline network according to the adjustment rules if the hydrogen sulfide content in the gas exceeds the standard. After the adjustment, the hydrogen sulfide content in the gas in the main pipeline network is calculated again. If the hydrogen sulfide content in the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content is adjusted again until the hydrogen sulfide content in the gas in the main pipeline network does not exceed the standard.

[0178] In the above implementation process, the system first obtains the training data collected from the DCS and preprocesses the training data to avoid the influence of null values ​​in the training data and keep the frequency of data acquisition consistent. Then, through correlation analysis and the least squares-based step-by-step method, parameters are screened from the preprocessed training data to form multiple structural equation models, and the model parameters of the optimal feature model are selected from multiple structural equation models and substituted into the prediction model. Then, the preprocessed training data is used to train the BP neural network model and the LSTM neural network model respectively, and the model is optimized from the trained BP neural network model and the trained LSTM neural network model to obtain the optimal prediction model. The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network, thereby predicting whether the SO2 exhaust gas of the heating furnace meets the standard. According to the regulation rules, the gas flow rate and amine liquid circulation volume are adjusted to make advance adjustments to the gas desulfurization system operating parameters, thereby ensuring that the SO2 emissions from the rear-end heating furnace exhaust meet the emission standards. The system promptly provides a continuous adjustment plan for optimizing the process parameters of the gas desulfurization system to ensure stable and standard SO2 emissions from the source.

[0179] Example 4

[0180] An embodiment of the present invention provides a machine-readable storage medium having instructions stored thereon. When the instructions are executed by the processor 100, the processor 100 is configured to execute the above-mentioned intelligent prediction and control method for a gas desulfurization system.

[0181] Machine-readable storage media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0182] The embodiment of the present application also provides an electronic device 10, which includes a memory 101, a processor 100, and a computer program 102 stored in the memory 101 and executable on the processor 100. When the processor 100 executes the computer program 102, the above-mentioned intelligent prediction and control method for the gas desulfurization system is implemented.

[0183] like Fig.10 FIG. 1 is a schematic diagram of an electronic device provided by an embodiment of the present invention. Fig.10 As shown, the electronic device 10 of this embodiment includes: a processor 100, a memory 101, and a computer program 102 stored in the memory 101 and executable on the processor 100. When the processor 100 executes the computer program 102, the steps in the above method embodiment are implemented. Alternatively, when the processor 100 executes the computer program 102, the functions of each module / unit in the above device embodiment are implemented.

[0184] Exemplarily, the computer program 102 may be divided into one or more modules / units, one or more modules / units are stored in the memory 101, and are executed by the processor 100 to implement the present invention. One or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 102 in the terminal device 10.

[0185] The electronic device 10 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 10 may include, but is not limited to, a processor 100 and a memory 101. Those skilled in the art will appreciate that Fig.10 It is only an example of the electronic device 10 and does not constitute a limitation of the electronic device 10. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0186] The processor 100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0187] The memory 101 may be an internal storage unit of the electronic device 10, such as a hard disk or memory of the electronic device 10. The memory 101 may also be an external storage device of the electronic device 10, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 10. Further, the memory 101 may also include both an internal storage unit of the electronic device 10 and an external storage device. The memory 101 is used to store computer programs and other programs and data required by the electronic device 10. The memory 101 may also be used to temporarily store data that has been output or is to be output.

[0188] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0189] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may 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 may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0190] 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 100 of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor 100 of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. 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.

[0191] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0193] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0194] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for intelligent prediction and control of a gas desulfurization system, characterized in that: include: Parameters are screened from the preprocessed training data to form multiple structural equation models; Determine an optimal feature model from multiple structural equation models, and substitute model parameters of the optimal feature model into the prediction model as modeling parameters incorporated into the prediction model, wherein the prediction model includes a BP neural network model and an LSTM neural network model; Use the preprocessed training data to train the BP neural network model and the LSTM neural network model respectively; According to the preset model evaluation index, the optimal prediction model is determined from the trained BP neural network model and the trained LSTM neural network model; The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network; If the hydrogen sulfide content of the gas in the main pipeline network exceeds the standard, the gas with excessive hydrogen sulfide content shall be adjusted according to the regulation rules, and the hydrogen sulfide content of the gas in the main pipeline network shall be calculated again after adjustment. If the hydrogen sulfide content of the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content shall be adjusted again until the hydrogen sulfide content of the gas in the main pipeline network does not exceed the standard.

2. The intelligent prediction and control method for gas desulfurization system according to claim 1 is characterized in that: Also includes: According to the preset data update frequency, the hydrogen sulfide content of the main pipeline network gas is calculated and updated in real time.

3. The intelligent prediction and control method for gas desulfurization system according to claim 2 is characterized in that: Also includes: Based on the delivery time of gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network, it is predicted whether the delivered gas will cause the SO2 at the outlet of the heating furnace device to exceed the standard.

4. The intelligent prediction and control method for gas desulfurization system according to claim 3 is characterized in that: The method further includes: predicting whether the transported gas will cause the outlet SO2 of the heating furnace device to exceed the standard based on the transport time of the gas to each heating furnace device and the real-time updated hydrogen sulfide content of the gas in the main pipeline network. According to the length of the gas pipeline network and the gas transmission flow rate, the formula Calculate the delivery time for gas to reach each heating furnace device, where t is the delivery time, L is the length of the gas pipe network, and v is the gas delivery flow rate.

5. The intelligent prediction and control method for gas desulfurization system according to claim 1 is characterized in that: The parameter screening from the preprocessed training data includes: Parameter screening is performed from the preprocessed training data through correlation analysis and least squares-based stepwise method.

6. The intelligent prediction and control method for gas desulfurization system according to claim 5 is characterized in that: The correlation analysis includes Pearson correlation analysis and Spearman correlation analysis; Before the parameter screening is performed from the pre-processed training data by correlation analysis and the least squares-based stepwise method, the method further includes: Perform normality analysis on the preprocessed training data; If the preprocessed training data conforms to the normal distribution, Pearson correlation analysis is used; If the preprocessed training data did not conform to a normal distribution, Spearman correlation analysis was used.

7. The intelligent prediction and control method for gas desulfurization system according to claim 5, characterized in that: The method of screening parameters from the preprocessed training data to form multiple structural equation models includes: Performing correlation analysis on the preprocessed training data to obtain correlation analysis results, and using a least squares-based stepwise method to select parameters from the preprocessed training data to obtain a least squares stepwise method analysis result; According to the correlation analysis results, the parameters that meet the correlation and collinearity requirements are screened and incorporated into the first structural equation model, and the included parameters are classified as latent variables according to pressure, flow rate, and temperature; According to the least squares stepwise analysis results, the model parameters that meet the requirements of multivariate and goodness of fit are screened and incorporated into the second structural equation model, and the included parameters are classified as latent variables according to pressure, flow, and hydraulic control; Combined with the correlation analysis results and the least squares stepwise analysis results, common parameters are screened and incorporated into the third structural equation model, and the incorporated parameters are classified as latent variables according to pressure, flow, and hydraulic control.

8. The intelligent prediction and control method for gas desulfurization system according to claim 1 is characterized in that: Also includes: Obtain training data collected from the DCS of the gas desulfurization system.

9. The intelligent prediction and control method for gas desulfurization system according to claim 8, characterized in that: Also includes: The training data is preprocessed.

10. The intelligent prediction and control method for gas desulfurization system according to claim 9, characterized in that: The preprocessing of the training data comprises: Perform outlier screening and difference processing on the training data and delete null value data; Resample the training data.

11. The intelligent prediction and control method for gas desulfurization system according to claim 8, characterized in that: The acquisition of training data collected from the DCS of the gas desulfurization system includes: According to the correlation between the SO2 data at the outlet of the heating furnace device and the control parameters of the hydrogen sulfide content in the gas, multiple influencing indicators for controlling the hydrogen sulfide content in the gas are determined; According to the multiple influencing indicators, training data collected from the DCS is acquired.

12. The intelligent prediction and control method for gas desulfurization system according to claim 1, characterized in that: The step of determining the optimal characteristic model from a plurality of structural equation models comprises: The chi-square freedom ratio, goodness-of-fit index, and root mean square error of approximation of each structural equation model were compared to determine the optimal characteristic model.

13. The intelligent prediction and control method for gas desulfurization system according to claim 1, characterized in that: The method of using the preprocessed training data to respectively train the BP neural network model and the LSTM neural network model includes: Divide the preprocessed training data into a training set and a test set according to a preset ratio; The training set is used to train the BP neural network model and the LSTM neural network model respectively, and the test set is used to verify the BP neural network model and the LSTM neural network model.

14. The intelligent prediction and control method for gas desulfurization system according to claim 1, characterized in that: The preset model evaluation indicators include root mean square error RMSE and determination coefficient R 2 .

15. The intelligent prediction and control method for gas desulfurization system according to claim 14, characterized in that: The method of determining the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index includes: The root mean square error RMSE and the coefficient of determination R 2 The neural network model with a higher value is taken as the optimal prediction model.

16. The intelligent prediction and control method for gas desulfurization system according to claim 1, characterized in that: The optimal prediction model is used to predict the hydrogen sulfide content of each gas stream, and the flow data of each gas stream is used as a weight to calculate the hydrogen sulfide content of the gas in the main pipeline network. The calculation formula is as follows: CS=Q1 / (Q1+Q2+…+Qn)×CS1+Q2 / (Q1+Q2+…+Qn)×CS2+…+Qn / (Q1+Q2+…+Qn)×CSn, wherein CS is the hydrogen sulfide content of the gas in the main pipeline network, CS1, CS2, ...CSn are the hydrogen sulfide contents of each gas stream, Q1, Q2, ...Qn are the flow rates of each gas stream, and Q1 / (Q1+Q2+…+Qn), ...Qn / (Q1+Q2+…+Qn)×CSn are the weight coefficients of each gas stream respectively.

17. The intelligent prediction and control method for gas desulfurization system according to claim 1, characterized in that: According to the regulation rules, the gas with excessive hydrogen sulfide content is regulated, including: Reduce the gas inlet flow rate of the gas desulfurization system and increase the amine liquid circulation volume of the gas desulfurization system in steps of 1%.

18. An intelligent prediction and control system for a gas desulfurization system, characterized in that: include: A parameter screening module is used to screen parameters from the preprocessed training data to form multiple structural equation models; An optimal feature model determination module, used to determine the optimal feature model from multiple structural equation models, and substitute the model parameters of the optimal feature model into the prediction model as modeling parameters incorporated into the prediction model, wherein the prediction model includes a BP neural network model and an LSTM neural network model; The model training module is used to train the BP neural network model and the LSTM neural network model respectively using the preprocessed training data; The optimal prediction model determination module is used to determine the optimal prediction model from the trained BP neural network model and the trained LSTM neural network model according to the preset model evaluation index; A hydrogen sulfide content prediction module is used to predict the hydrogen sulfide content of each gas stream by using the optimal prediction model, and to calculate the hydrogen sulfide content of the gas in the main pipeline network by using the flow data of each gas stream as a weight; The gas adjustment module is used to adjust the gas with excessive hydrogen sulfide content in the main pipeline network according to the adjustment rules if the hydrogen sulfide content in the gas exceeds the standard. After the adjustment, the hydrogen sulfide content in the gas in the main pipeline network is calculated again. If the hydrogen sulfide content in the gas in the main pipeline network still exceeds the standard, the gas with excessive hydrogen sulfide content is adjusted again until the hydrogen sulfide content in the gas in the main pipeline network does not exceed the standard.

19. A machine-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by a processor, the processor is configured to execute the intelligent prediction and control method for a gas desulfurization system as described in any one of claims 1 to 17.

20. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the intelligent prediction and control method for the gas desulfurization system according to any one of claims 1 to 17 is implemented.

Images