Cracking furnace real-time production optimization method, electronic equipment and storage medium

Abstract

The invention discloses a real-time production optimization method for a cracking furnace, electronic equipment and a storage medium. The method comprises the following steps: collecting and preprocessing historical operation data of the cracking furnace to obtain key data; analyzing and establishing a target cracking depth model according to the key data; optimizing the key data according to the target cracking depth model to obtain a production target function; acquiring real-time operation data of the cracking furnace; and optimizing the production of the cracking furnace in real time according to the production objective function and the real-time operation data. Therefore, the production income of the cracking furnace is improved, the operation condition of the cracking furnace is monitored in real time, and operation faults are predicted in advance.

Description

Technical Field

[0001] The present invention relates to the technical field related to chemical production, in particular to a real-time production optimization method for a cracking furnace, an electronic device, and a storage medium. Background Art

[0002] The ethylene plant is the leading part in the chemical production process. By subjecting raw materials such as light hydrocarbons, naphtha, and hydrocracking tail oil to cracking reactions in a cracking furnace under high-temperature conditions, important chemical products such as ethylene and propylene, as well as high-value-added products such as hydrogen and aromatics, are produced. The cracking furnace is an important place where the cracking reaction occurs and is also the main energy consumption source of the ethylene plant. At the same time, in addition to the main and side cracking reactions, a series of processes such as flow, mixing, mass transfer, heat transfer, and the coupling processes of such processes also occur in the cracking furnace, and there is still some room for improvement in the adequacy of the research on its mechanism. Moreover, compared with raw materials such as naphtha and hydrocracking tail oil, light hydrocarbon raw materials such as propane and ethane have more stringent requirements and controls for operating conditions. Therefore, building a model of an ethylene light hydrocarbon cracking furnace is of great guiding significance for its operation optimization, product revenue improvement, energy conservation, and consumption reduction.

[0003] Currently, the optimization of chemical processes mostly takes an offline form, that is, modeling and analyzing a certain unit, process, or operation, etc. by means of historical operation data under certain working conditions. However, due to the influence of factors such as changes in operating conditions, degradation of production equipment performance, and adjustment of production policies, the offline optimization process often has a certain lag in guiding production operations, and there is a certain deviation between the optimization calculation results and expectations. In addition, the offline optimization results often only consider the optimization process itself, cannot timely capture the changes in equipment performance and surrounding environmental factors, and cannot timely reflect the impact of such changes on the optimization process and results in the optimization model. Therefore, the implementation of real-time online modeling and optimization is of great significance.

[0004] Previous optimization research on ethylene cracking furnaces has shown four characteristics: First, regarding the research on the cracking reaction mechanism inside the cracking furnace, this part of the work is highly complex, has a long research cycle, and there are many ideal factors in its results, making it unable to match well with the actual operating conditions. Second, simple optimization at the operation level, such as replacing the furnace tube material and surface coating material, raising or lowering the overall furnace COT based on personal experience, etc., has mixed results in terms of the optimization effect. Third, through the analysis of the historical operation data of the cracking furnace and building its data model through a simple algorithm, there is no qualitative or quantitative adjustment basis for the deviation between the model and the actual value, so the utilization rate of the model is poor. Fourth, most cracking furnaces in domestic and foreign ethylene plants are liquid-phase furnaces, usually using heavy oil products such as naphtha and hydrocracked tail oil as raw materials, with relatively rich operation experience, while gas-phase cracking furnaces using light hydrocarbons such as propane and ethane as raw materials are relatively few. Considering the differences in furnace structure and operating conditions, etc., the cracking reactions occurring inside the furnace are not exactly the same. At the same time, traditional optimization work only stays at offline adjustment. There are often certain changes in the operating conditions corresponding to the data used for optimization calculation and the operating conditions corresponding to the optimization result output. The optimization result using offline data has poor timeliness for production guidance, resulting in production not achieving the best benefits and failing to achieve real-time monitoring of the operating status and unable to predict operating failures in advance. Summary of the Invention

[0005] Based on this, it is necessary to provide a real-time production optimization method, electronic device, and storage medium for a cracking furnace to address technical problems such as not achieving the best benefits due to optimizing production using offline data, failing to achieve real-time monitoring of the operating status, and being unable to predict operating failures in advance.

[0006] The present invention provides a real-time production optimization method for a cracking furnace, including:

[0007] Collect and preprocess the historical operation data of the cracking furnace to obtain key data;

[0008] Establish a target cracking depth model based on the analysis of the key data;

[0009] Optimize the key data according to the target cracking depth model to obtain a production objective function;

[0010] Obtain the real-time operation data of the cracking furnace;

[0011] Real-time optimize the production of the cracking furnace according to the production objective function and the real-time operation data.

[0012] Further, the collecting and preprocessing the historical operation data of the cracking furnace to obtain key data includes:

[0013] Collect and organize the historical operation data, and use the data smoothing method to clean the historical operation data to obtain the cleaned historical operation data;

[0014] Conduct a correlation analysis on the cleaned historical operation data to obtain key data.

[0015] Furthermore, establishing the target cracking depth model based on the analysis of the key data includes:

[0016] Use the neural network algorithm to model the key data to obtain the initial cracking depth model;

[0017] Obtain the model cracking depth and the actual cracking depth of the initial cracking depth model;

[0018] Analyze and obtain the cracking depth deviation based on the model cracking depth and the actual cracking depth;

[0019] Iteratively optimize the initial cracking depth model according to the cracking depth deviation to obtain the target cracking depth model.

[0020] Furthermore, the key data includes production operation data, production revenue data, and production consumption data. The production objective function includes a production revenue objective function and a production consumption objective function. Optimizing the key data according to the target cracking depth model to obtain the production objective function includes:

[0021] Analyze and optimize the production revenue data according to the production operation data and the target cracking depth model to obtain the production revenue objective function;

[0022] Analyze and optimize to obtain the production consumption objective function according to the production operation data and the production consumption data.

[0023] Furthermore, optimizing the production of the cracking furnace in real time according to the production objective function and the real-time operation data includes:

[0024] Deduce the production operation data according to the preset production objective and the production objective function;

[0025] Determine the operating state of the cracking furnace according to the real-time operation data;

[0026] Determine whether there are any potential operating hazards in the cracking furnace according to the real-time operation data.

[0027] Furthermore, deducing the production operation data according to the preset production objective and the production objective function includes:

[0028] Input the production target as the dependent variable into the production target function to infer the production operation data.

[0029] Further, determining the operating state of the cracking furnace according to the real-time operation data includes:

[0030] Calculate the thermal efficiency from the real-time operation data;

[0031] Judge whether the thermal efficiency is less than the preset thermal efficiency value. If the thermal efficiency is less than the preset thermal efficiency value, prompt that the cracking furnace needs to be overhauled. If the thermal efficiency is greater than or equal to the preset thermal efficiency value, prompt that the operating state is normal.

[0032] Further, determining whether there are potential operating hazards in the cracking furnace according to the real-time operation data includes:

[0033] Set the standard range of the key data;

[0034] If the real-time operation data is not within the standard range, issue a warning of the potential operating hazard.

[0035] The present invention provides an electronic device, including:

[0036] At least one processor; and,

[0037] A memory communicatively connected to at least one of the processors; wherein,

[0038] The memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute the cracking furnace real-time production optimization method as described above.

[0039] The present invention provides a storage medium that stores computer instructions, and when a computer executes the computer instructions, it is used to execute all steps of the cracking furnace real-time production optimization method as described above.

[0040] The present invention analyzes key data based on the historical operation data of the cracking furnace, establishes a cracking depth model based on the key data, optimizes the key data according to the model to obtain a target production function, and at the same time realizes real-time monitoring and optimization of the cracking furnace according to the production target function and real-time operation data, thereby achieving the improvement of the production benefit of the cracking furnace, realizing the real-time monitoring of the operation status of the cracking furnace, and predicting operation failures in advance. Description of the Drawings

[0041] Figure 1 It is a working flow chart of a cracking furnace real-time production optimization method according to an embodiment of the present invention;

[0042] Figure 2 This is the flowchart of a real-time production optimization method for a cracking furnace according to another embodiment of the present invention;

[0043] Figure 3 This is the fuel gas consumption diagram before the improvement of the multi-objective genetic optimization algorithm according to an embodiment of the present invention;

[0044] Figure 4 This is the fuel gas consumption diagram after the improvement of the multi-objective genetic optimization algorithm according to an embodiment of the present invention;

[0045] Figure 5 This is the schematic diagram of the hardware structure of an electronic device according to the present invention. Detailed implementation manners

[0046] The following further describes the detailed implementation manners of the present invention with reference to the drawings. The same components are denoted by the same reference numerals. It should be noted that the terms "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to the directions in the drawings, and the terms "inner" and "outer" respectively refer to the directions towards or away from the geometric center of a specific component.

[0047] As Figure 1 shown, this is the flowchart of a real-time production optimization method for a cracking furnace according to an embodiment of the present invention, including:

[0048] Step S101: Collect and preprocess the historical operation data of the cracking furnace to obtain key data;

[0049] Step S102: Analyze and establish a target cracking depth model based on the key data;

[0050] Step S103: Optimize the key data according to the target cracking depth model to obtain a production objective function;

[0051] Step S104: Obtain the real-time operation data of the cracking furnace;

[0052] Step S105: Optimize the production of the cracking furnace in real time according to the production objective function and the real-time operation data.

[0053] Specifically, in step S101, a large amount of historical operation data of the cracking furnace is collected first, and then these historical operation data are preprocessed such as cleaning. After analysis, key data is obtained. The key data includes production operation data, production revenue data, and production consumption data. Then, in step S102, the optimized neural network algorithm is used to model the key data to obtain the initial cracking depth model. The cracking depth deviation is obtained by comparing the model cracking depth and the actual cracking depth of the initial cracking depth model. The initial cracking depth model is optimized according to the cracking depth deviation to obtain the target cracking depth model. Then, in step S103, the improved multi-objective optimization genetic algorithm is used to simultaneously optimize the production revenue data and the production consumption data according to the target cracking depth model to obtain the production objective function. The production objective function includes the production revenue objective function and the production consumption objective function. Then, in step S104, the real-time operation data of the cracking furnace is acquired through collection. The real-time operation data is the key data when the cracking furnace is running in real time. Finally, in step S105, the production operation data is inferred according to the production objective function to guide the production operation. At the same time, the operation state of the cracking furnace, the real-time monitoring instrument failure, and the real-time optimization of the process are determined according to the real-time operation data.

[0054] In the present invention, key data is obtained after data processing of the historical operation data of the cracking furnace, and then the target cracking depth model is established by analyzing the key data. Then, multi-objective optimization is performed on the model to obtain the production objective function. The production of the cracking furnace is optimized in real time according to the production objective function and the acquired real-time operation data. Thereby, the production revenue of the cracking furnace is increased, the operation status of the cracking furnace is monitored in real time, and the operation failure is predicted in advance.

[0055] As Figure 2 shown is the working flowchart of a real-time production optimization method for a cracking furnace in another embodiment of the present invention, including:

[0056] Step S201: Collect and sort out the historical operation data, and use the data smoothing method to clean the historical operation data to obtain the cleaned historical operation data; perform correlation analysis on the cleaned historical operation data to obtain key data.

[0057] Specifically, the historical operation data of the cracking furnace in different cycles are collected, sorted, and processed by the multi-point averaging method and the five-point cubic smoothing method to eliminate signal burrs and improve data quality, obtaining the processed historical operation data. Then, through Pearson correlation analysis, variables linearly related to the cracking depth of the cracking furnace can be screened out, that is, the key data are obtained. At the same time, considering the complexity of the chemical production process and the non-linear characteristics of the cracking reaction in the cracking furnace, based on Pearson correlation analysis, variables non-linearly related to the cracking depth of the cracking furnace are determined by the maximum mutual information entropy analysis method to ensure the comprehensiveness and effectiveness of the data used for cracking depth modeling.

[0058] Specifically, the data of the cracking furnace in different operation cycles are processed by the multi-point averaging method and the five-point cubic smoothing method. Specifically, it includes: first performing multi-point averaging on the historical operation data of the cracking furnace to obtain a series of average value points: x1, x2, x3…x n . Where n is the number of average value points. The obtained average value points are subjected to five-point cubic smoothing processing, and a cubic polynomial Y = a0 + a1x1 + a2x2 + a3x3 is used for approximation during the smoothing process. Among them, the coefficients a1, a2, and a3 in the cubic polynomial are mainly determined by the least squares principle. On the basis of data processing, Pearson correlation analysis and the maximum mutual information entropy method are successively used to screen and extract the variables affecting the cracking depth of the cracking furnace.

[0059] The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables. By estimating the covariance and the standard deviation of the sample, the Pearson correlation coefficient of any two sample variables can be obtained, and the Pearson correlation coefficient can also be estimated from the mean of the standard scores of the sample points. The value range of the Pearson correlation coefficient is [-1, 1]. The larger its absolute value, the stronger the correlation between the two sample variables, and the smaller the absolute value, the weaker the correlation; a negative number indicates a negative correlation between variables, and a positive number indicates a positive correlation between variables.

[0060] On the basis of Pearson correlation analysis, maximum mutual information entropy analysis is performed to further screen variables non-linearly correlated with the cracking depth. Mutual information entropy is the difference in information entropy between two random variables. The larger the mutual information entropy value, the stronger the correlation between the two variables, and vice versa.

[0061] The processed historical operation data are successively subjected to Pearson correlation analysis and the maximum mutual information entropy method to screen and extract the variables affecting the cracking depth of the cracking furnace. Specifically, it includes:

[0062] The Pearson correlation coefficient between two variables is defined as the quotient of the covariance and the standard deviation between the two variables, and the calculation formula is:

[0063]

[0064] where ρ X，Y is the Pearson correlation coefficient of variable X and variable Y, cov(X, Y) is the covariance of X and Y, and σ X is the sample standard deviation of X, and σ Y is the sample standard deviation of Y, μ X is the population mean of X, and μ Y is the population mean of Y, and E[(X - μ X )(Y - μ Y )] represents the covariance of X and Y.

[0065] Estimate the covariance and standard deviation of the samples to obtain the Pearson correlation coefficient γ of any two sample variables. The calculation formula is:

[0066]

[0067] Alternatively, the Pearson correlation coefficient γ of any two sample variables is estimated by the mean of the standard scores of the sample points. The calculation formula is:

[0068]

[0069] where X i is the sample point of X, Y i is the sample point of Y, is the sample mean of X, is the sample mean of Y, represents the sample standard score of X, represents the sample standard score of Y.

[0070] Step S202: Use the neural network algorithm to model the key data to obtain an initial cracking depth model; obtain the model cracking depth and the actual cracking depth of the initial cracking depth model; analyze the cracking depth deviation based on the model cracking depth and the actual cracking depth; and iteratively optimize the initial cracking depth model according to the cracking depth deviation to obtain the target cracking depth model.

[0071] Specifically, the cracking depth is a parameter characterizing the degree of the cracking reaction. The higher the cracking depth, the higher the conversion rate and the greater the amount of gas-phase products. Here, the model cracking depth refers to the cracking depth calculated by this model, and the actual cracking depth refers to the cracking depth in actual production under the same working conditions simulated by this model. The initial cracking depth model is optimized so that the model cracking depth approaches the actual cracking depth to meet the requirements of real-time production. At this time, the target cracking depth model is obtained. After obtaining the key data linearly and non-linearly related to the cracking depth, with the help of the improved neural network algorithm IMBP, based on the traditional neural network algorithm, by introducing the shape parameter α and the fluctuation tolerance parameter i into the built-in general activation function Sigmoid, the initial cracking depth model is obtained; considering the complexity of the reaction, flow, mixing, mass transfer, heat transfer and other processes and their coupling processes in the cracking furnace, the deviation between the model cracking depth and the actual cracking depth of the initial cracking depth model is obtained again with the help of the improved neural network algorithm IMBP, and the deviation is modeled and analyzed with the variables affecting the cracking depth to obtain the cracking depth deviation; the cracking depth deviation is introduced into the initial cracking depth model to perform iterative correction until the deviation meets the process requirements. At this time, the target cracking depth model is obtained.

[0072] In some embodiments, after the topological structure and training data of the neural network are determined, the accuracy of the initial cracking depth model mainly depends on the choice of the activation function f to a certain extent. The influence of the f function on the model accuracy lies in reducing the possibility of local minima by changing the error surface. The f function adopts the form of the activation function Sigmoid, as follows:

[0073] f(x) = 1 / (1 + e^(-x))

[0074] Among them, the activation function Sigmoid is a saturated non-linear function with a saturation region. When the output of the neuron falls into the saturation region of the activation function, a large correction needs to be made to the weight value to make the processing unit escape the saturation region as soon as possible. Since in the saturation region, the derivative values of the activation function are all very small, each learning cycle can only make a small correction to the weight value, and the output unit will work in the flat region for a period of time, keeping the mean square error of the network unchanged or changing very little, thus slowing down the convergence speed of the network. The adopted activation function is as follows:

[0075] f(x) = 1 / (2N + 1) * 1 / (1 + e^(-α(x + i)))

[0076] Among them, x represents the independent variable of the activation function; N represents the number of independent variables; i represents the fluctuation tolerance parameter of the independent variable, that is, half of the absolute value of the fluctuation range, and its value is related to the fluctuation of variables under different working conditions; α is the shape parameter. When α decreases, the function shape is stretched horizontally, otherwise it is compressed horizontally, and the value of α is between 0 and 1.

[0077] By introducing the i and α parameters, the incentive function can more fully express the characteristics of the variables used for modeling, effectively accelerating the convergence rate of the initial cracking depth model, improving the accuracy of the initial cracking depth model, and finally obtaining the target cracking depth model.

[0078] Step S203, the key data includes production operation data, production revenue data, and production consumption data. The production objective function includes a production revenue objective function and a production consumption objective function. Optimizing the key data according to the target cracking depth model to obtain the production objective function includes: analyzing and optimizing the production revenue data according to the production operation data and the target cracking depth model to obtain the production revenue objective function; analyzing and optimizing to obtain the production consumption objective function according to the production operation data and the production consumption data.

[0079] Specifically, the production revenue objective function and the production consumption objective function refer to the multi-objective optimization result that maximizes production revenue and minimizes production consumption. Through the improved multi-objective optimization genetic algorithm IMNSGA_II, within a certain range of operating condition constraints, real-time optimization calculations are performed on the production operation data corresponding to the better production revenue data and production consumption data of the cracking furnace to obtain the production revenue objective function and the production consumption function.

[0080] In some embodiments, the production consumption data is the fuel gas consumption of the cracking furnace.

[0081] In the present invention, the multi-objective genetic optimization algorithm NSGA_II adopts an improved specific crossover operator combined with the actual production process, and its improved form is as follows:

[0082]

[0083]

[0084] Among them, and are respectively the true coding values of the decision variables corresponding to the two individual crossover points a and b in the t-th generation; m is a constant.

[0085] The cracking depth is an important characterization of the composition of cracked gas. For the analysis of the composition of the cracked gas at the outlet of the cracking furnace, conventional chromatographs can usually only analyze components such as hydrogen, methane, acetylene, ethylene, ethane, MA, PD, propylene, and propane, and the composition of components with four or more carbon atoms cannot be obtained. Therefore, the sum value of the on-line analytical instruments is usually <100%. At the same time, considering the deviation between the target cracking depth model and the actual cracking depth, when the calculated value of the target cracking depth model is greater than the actual value, the value of m is set as the sum value of the key component contents of the on-line analytical instruments / 100%; when the model calculated value is less than or equal to the actual value, the value of m is 100% / the sum value of the key component contents of the on-line analytical instruments.

[0086] Based on the target cracking depth model, the present invention performs dual optimization calculations for maximizing the product revenue of the cracking furnace and minimizing the fuel gas consumption by means of the improved multi-objective genetic optimization algorithm IMNSGA_II; by specifically modifying the constant coefficient m in the individual crossover calculation process in combination with the sum value data of the key product components of the on-line analytical instrument at the outlet of the cracking furnace, specific optimization of the binary crossover operator SBX is realized, and thus the multi-objective optimization calculation results are closer to the actual production process and more feasible and implementable.

[0087] The form and constraint conditions of the multi-objective optimization function of the cracking furnace are determined as follows:

[0088]

[0089] Zmin = FGM

[0090] Q min ≤Q≤Q max

[0091] COT min ≤COT≤COT max

[0092] DSR min ≤DSR≤DSR max

[0093] TXT min ≤TXT≤TXT max

[0094] PR ≤ 0.9

[0095] Among them, Ymax is the maximum value of the product revenue of the cracking furnace; Zmin is the minimum value of the fuel gas consumption of the cracking furnace; the value range of i is one of 1, 2, 3, 4, 5, 6, representing six key product components: hydrogen, methane, ethylene, ethane, propylene, and propane; the value of n is 6; PM represents the product output; P represents the price; FGM represents the fuel gas consumption; Q represents the feed flow rate of the cracking furnace, Q min and Qmax respectively represent the minimum and maximum values of the feed flow rate of the cracking furnace; COT represents the outlet temperature of the cracking furnace, COT min and COT max respectively represent the minimum and maximum values of the outlet temperature of the cracking furnace; DSR represents the dilution steam ratio, DSR min and DSR max respectively represent the minimum and maximum values of the dilution steam ratio; TXT represents the feed temperature across the section, TXT min and TXT max respectively represent the minimum and maximum values of the feed temperature across the section; PR represents the absolute pressure ratio of the furnace tube.

[0096] In the present invention, through the improved multi-objective optimization genetic algorithm IMNSGA_II, within a certain range of operating condition constraints, simultaneous optimization calculations are performed on two objective functions of the cracking furnace product yield and fuel gas consumption. By optimizing m, the search ability of the target cracking depth model is improved. While ensuring the population diversity, the search area of the target cracking depth model is further expanded, and thus the result of the target cracking depth model is closer to the actual production process.

[0097] Step S204, obtaining the real-time operation data of the cracking furnace.

[0098] Specifically, the real-time operation data read mainly includes: the feed flow rate of the cracking furnace, the outlet temperature of the cracking furnace (COT), the outlet pressure of the cracking furnace (COP), the dilution steam ratio (DSR), the feed temperature across the section, the absolute pressure ratio (PR), and the upper and lower limit data of the above key data, the on-line instrument composition at the outlet of the cracking furnace (including components such as hydrogen, methane, ethylene, ethane, propylene, propane, etc.) and the price system data. Among them, the price data mainly includes the prices of various products and raw materials. At the same time, through the calculation of the read real-time operation data, the product components, the cracking furnace product yield, the fuel gas consumption of the cracking furnace, and the cracking depth model deviation data are obtained. Therefore, part of the real-time operation data is obtained by direct reading, and the other part is obtained through calculation.

[0099] Step S205, inferring production operation data according to the preset production target and the production target function; determining the operation state of the cracking furnace according to the real-time operation data; determining whether there are operation hazards of the cracking furnace according to the real-time operation data.

[0100] In one embodiment, the inferring production operation data according to the preset production target and the production target function includes: inputting the production target as the dependent variable into the production target function to infer the production operation data.

[0101] Specifically, based on the target cracking depth model, the key data is optimized by combining the real-time operation data and the production target to obtain the optimal production operation data. The production target is preset according to the production conditions and requirements, and then the production target is used as the dependent variable to input into the production target function to calculate and infer the production operation data, guiding the production operation, so as to achieve the preset production target.

[0102] In one embodiment, determining the operating state of the cracking furnace according to the real-time operation data includes:

[0103] Calculating the thermal efficiency from the real-time operation data;

[0104] Judging whether the thermal efficiency is less than the preset value of the thermal efficiency. If the thermal efficiency is less than the preset value of the thermal efficiency, it is prompted that the cracking furnace needs to be overhauled. If the thermal efficiency is greater than or equal to the preset value of the thermal efficiency, it is prompted that the operating state is normal.

[0105] Specifically, the monitoring variables of the cracking furnace mainly include the composition content data of key components of the on-line analysis instrument at the outlet of the cracking furnace, the flue gas temperature of the cracking furnace, the absolute pressure ratio, the operating time of the cracking furnace and the expected operating time, etc. Among them, the flue gas temperature of the cracking furnace is mainly used for calculating the thermal efficiency of the cracking furnace. The specific calculation formula is: thermal efficiency = A * flue gas temperature of the cracking furnace + B, where both A and B are constant coefficients. Finally, the operating state of the cracking furnace is judged according to the thermal efficiency. If the thermal efficiency is less than the preset value of the thermal efficiency, it is prompted to overhaul the cracking furnace. If the thermal efficiency is greater than or equal to the preset value of the thermal efficiency, it is prompted that the operating state is normal.

[0106] This embodiment realizes real-time monitoring of the operating state of the cracking furnace according to the thermal efficiency.

[0107] In one embodiment, determining whether there are potential operating hazards of the cracking furnace according to the real-time operation data includes:

[0108] Setting the standard range of the key data;

[0109] If the real-time operation data is not within the standard range, an early warning of the potential operating hazard is issued.

[0110] Specifically, through the accumulation of production experience of the cracking furnace for a period of time, the standard range of the key data of the cracking furnace can be obtained. When it is detected that the real-time operation data is abnormal and out of the standard range, an early warning of the corresponding potential operating hazard is timely sent to the operator to remind the operator to check and correct the corresponding problems.

[0111] In some embodiments, when there are no fluctuations or abnormalities in the process conditions, the numerical value of the cracking furnace outlet temperature (COT) generally does not exceed this standard range for operation. It should be particularly noted that the high-temperature environment inside the cracking furnace usually causes failures in some measuring instruments, such as component damage or zero drift. When the COT data exceeds this standard range, functions such as color mutation, flashing, and alarm are used to prompt the operator to pay attention in a timely manner, and to conduct cause investigation, analysis, and correction.

[0112] The on-line analyzer at the cracking furnace outlet can measure and display the composition of key products after the cracking reaction, such as hydrogen, methane, ethylene, ethane, propylene, propane, etc. In the industry, chromatography is generally used. Since the cracking reaction produces not only light hydrocarbon products such as hydrogen - propane, but also heavier liquid oil components such as mixed C4 and pyrolysis gasoline. If such components are not completely condensed in the pretreatment system of the chromatography, they will enter the chromatography for composition analysis along with the above-mentioned light hydrocarbon components, which will cause pollution of the chromatographic column and blockage of the chromatographic injector. As a result, the composition analysis of the chromatography, that is, the display data of the on-line analyzer, will be abnormal. Common abnormal readings mainly include: the sum value of key component data is less than 95%, a certain component shows a straight line, or the data of a certain component suddenly jumps and fluctuates during normal operation of the process. Once such data anomalies are found, the operator is promptly prompted to analyze and solve the problems.

[0113] Steady-state detection is to judge whether the cracking furnace is operating in a stable state. The numerical value of the steady-state detection result of the cracking furnace needs to be greater than or equal to 85% to indicate that the cracking furnace is operating stably. Usually, assume that the steady-state detection result of a furnace is 90%, and it drops to 85% during a certain period. Although it is still in a stable operating state, compared with the previous result value, it has decreased. At this time, if the process conditions have not changed significantly, the decrease in the steady-state detection result is very likely related to the measuring instruments used for steady-state calculation. For the steady-state detection of the cracking furnace, three operating parameters, namely COT, cracking furnace feed flow rate, and DSR, are usually selected, and instrument problems are checked one by one according to these three operating parameters.

[0114] The absolute pressure ratio refers to the ratio of the absolute pressures before and after the venturi of the furnace tube, which is an important indicator parameter for determining whether to carry out decoking treatment on the cracking furnace. Generally speaking, the value range of the absolute pressure ratio is between 0.4 and 0.9, and the absolute pressure ratio values are different among different furnace tubes. When the absolute pressure ratio of any furnace tube in the furnace exceeds or equals 0.9, preparations for furnace decoking work need to be carried out. During a certain operation cycle, the absolute pressure ratio of the furnace tubes basically shows a certain regular growth trend. If, during a certain period of a certain operation cycle, the absolute pressure ratio of a furnace tube shows a sudden increase, based on production operation experience, it can generally be determined that it is caused by coke shedding, that is, the coke blocks accumulated by the cracking reaction fall from a higher position of the furnace tube to a lower position. Once the trend of the absolute pressure ratio changes abnormally, it prompts the operator to check the parameters of the furnace tube temperature equalization control loop, reset reasonable controller parameters according to the temperature difference data of each furnace tube to ensure that the material flow rate in each furnace tube is uniform and constant, or conduct timely and comprehensive inspection and analysis on other relevant on-site hardware facilities according to production experience to ensure that the cracking furnace is in a safe and stable operating state.

[0115] In this embodiment, by comparing the real-time operation data with the set standard range, the abnormality of the real-time operation data is timely detected, so as to timely issue a warning to the operator and timely eliminate operation hazards.

[0116] Optimal Embodiment

[0117] Taking the ethylene light hydrocarbon cracking furnace of a certain chemical enterprise as the research object, this cracking furnace has a single furnace structure, uses propane as the cracking raw material, and its cracking depth is represented by the propane conversion rate. The cracking depth data is mainly determined by the online analysis instrument data at the cracking furnace outlet and the propane content data in the feed. The historical operation data of this cracking furnace under different cycles since startup has been collected and sorted, a cracking depth model has been built, and multi-objective optimization has been carried out with maximizing production benefits and minimizing fuel gas consumption as the goals. At the same time, an online optimization calculation platform and a monitoring platform have been developed. The online (i.e., real-time) target cracking depth model and the online operation monitoring platform have been built through the computer programming language Python.

[0118] After collecting and organizing the historical operation data of a single furnace and conducting Pearson correlation analysis and maximum mutual information entropy analysis, the variables affecting the cracking depth in the furnace are mainly: feed rate, cracking furnace outlet temperature (COT), cracking furnace outlet pressure (COP), dilution steam ratio (DSR), and cross-section temperature (TXT). With the improved neural network algorithm IMBP provided by the present invention, a model of the cracking depth in the furnace is established to obtain an initial cracking depth model. At the same time, the cracking depth calculated by the initial cracking depth model under different operating conditions is compared with the actual cracking depth for analysis to obtain the cracking depth deviation. Then, the deviation and the variables affecting the cracking depth are modeled and analyzed using IMBP, and the deviation is introduced into the initial cracking depth model for iterative correction to make the deviation between the cracking depth calculated by the model and the actual data meet the process requirements, with an absolute deviation of ±0.2%. The comparison of the model calculation and the actual value before and after the correction of the neural network algorithm and before and after the iterative correction is shown in Table 1. Through the improved neural network algorithm with the introduction of the cracking depth deviation for iterative correction, a target cracking depth model is obtained. The cracking depth obtained from the target cracking depth model is closer to the actual value, basically meeting the process usage requirements.

[0119] Table 1

[0120]

[0121] A certain cracking depth corresponds to a certain product distribution and fuel gas consumption. Based on the target cracking depth model, optimization calculations are carried out for two objectives: product revenue (i.e., production revenue) and fuel gas consumption (i.e., production consumption). The calculation formula for the product revenue of the cracking furnace is defined as: the sum of the product quantities of hydrogen / methane / ethylene / ethane / propylene / propane / mixed C4 / cracked gasoline multiplied by their respective prices minus the propane raw material usage multiplied by the propane raw material price. Among them, the price data is sourced from the monthly financial settlement data of the enterprise. The main constraints for the multi-objective optimization calculation are: feed rate (36 - 46 t / h), COT (840 - 856 °C), COP (1.45 - 1.65 atm), cross-section temperature (550 - 650 °C), and absolute pressure ratio within 0.9. The calculation results after 100 iterations before and after the improvement of the multi-objective genetic optimization algorithm are shown in Figure 3 and Figure 4 respectively. By comparing the result graphs before and after the improvement, it can be seen that the curve after the improvement is smoother and the distribution of the optimization calculation points is more uniform, with a richer diversity of solutions.

[0122] Based on the multi-objective offline optimization of the cracking furnace, a communication interface connecting the target cracking depth model written on the Python platform with the real-time database PI (Plant Information System) is developed to obtain production operation data such as feed rate, COT, COP, DSR, and cross-section temperature in real time. To ensure the feasibility of the calculation results of the target cracking depth model, before introducing the real-time operation data into the model, the steady-state detection of the cracking furnace operation status is carried out by means of T-test (using the t-distribution theory to infer the probability of differences occurring and comparing whether the differences between two averages are significant) and F-test (joint hypothesis test). The steady-state detection variables of the cracking furnace are selected as feed rate, COT, and DSR, and their fluctuation tolerance parameters are ±2 t / h, ±2.5 °C, and ±0.01 respectively. The overall acceptable steady-state value of the cracking furnace is 85%, that is, when and only when 85 out of 100 points detected for each variable are within the fluctuation range among the feed rate, COT, and DSR variables, the cracking furnace operation status is in a steady state; otherwise, it is in an unsteady state. The overall steady-state value of the cracking furnace has only two values, 0 and 1, where 1 represents a steady state and 0 represents an unsteady state.

[0123] While performing multi-objective online optimization calculations, real-time monitoring is carried out on the status of the on-line analysis instrument at the outlet of the cracking furnace, the absolute pressure ratio, the flue gas temperature, the operating time, and the estimated operating time. Among them, the status of the on-line analysis instrument is mainly characterized by the sum of the components from hydrogen to propane and the relative deviation of the measured values of each component within one week. Under certain operating conditions, when the sum of the components from hydrogen to propane is 95% or more and the relative deviation values of hydrogen, methane, ethylene, ethane, propylene, and propane components in two adjacent cycles are respectively less than or equal to 2.5%, less than or equal to 3.5%, less than or equal to 5%, less than or equal to 3%, less than or equal to 3.5%, and less than or equal to 4%, it indicates that the measured values of the on-line analysis instrument are normal, and its relative deviation value is usually determined by the accuracy of the instrument itself and the changes in operating conditions; the device judges whether it is necessary to carry out coke burning treatment on the furnace chamber according to the value of the absolute pressure ratio. Usually, when the absolute pressure ratio is greater than or equal to 0.9, coke burning treatment is carried out; the thermal efficiency of the furnace chamber can be estimated through the flue gas temperature, and the specific calculation formula is: Furnace thermal efficiency,% = -0.0433 * Flue gas temperature + 98.84; The determination of the operating time of the cracking furnace is related to the judgment of its operating state: when the feed rate, fuel gas consumption, and dilution steam consumption are all ≥ 3 t / h, it indicates that the cracking furnace is in the operating state; when the feed, fuel gas consumption, and dilution steam consumption are all ≤ 3 t / h, it is in the shutdown state; when the feed and dilution steam consumption ≤ 3 t / h and the fuel gas consumption ≥ 3 t / h, it is in the equipment state; when the feed and fuel gas consumption ≤ 3 t / h and the dilution steam consumption ≥ 3 t / h, it is in the coke burning state. The operating time of the furnace chamber in the latest operating cycle is judged in turn according to these 4 states; the estimated operating time of the cracking furnace is related to the value of the absolute pressure ratio, and its calculation formula is: (0.9 - Current absolute pressure ratio) / (Rate of change of absolute pressure ratio).

[0124] As Figure 5 shown is a schematic hardware structure diagram of an electronic device according to the present invention, including:

[0125] At least one processor 501; and,

[0126] A memory 502 communicatively connected to at least one of the processors 501; wherein,

[0127] The memory 502 stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors so that at least one of the processors can execute the cracking furnace real-time production optimization method as described above.

[0128] Figure 5 One processor 501 is taken as an example in the above.

[0129] The electronic device may further include: an input device 503 and a display device 504.

[0130] The processor 501, the memory 502, the input device 503, and the display device 504 may be connected through a bus or other means. In the figure, the connection through the bus is taken as an example.

[0131] The memory 502, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the real-time production optimization method of the cracking furnace in the embodiments of the present application. For example, Figure 1 and Figure 2 the method flow shown. The processor 501 executes various functional applications and data processing by running the non-volatile software programs, instructions, and modules stored in the memory 502, that is, realizes the real-time production optimization method of the cracking furnace in the above embodiments.

[0132] The memory 502 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the real-time production optimization method of the cracking furnace, etc. In addition, the memory 502 may include a high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 502 may optionally include a memory remotely set relative to the processor 501, and these remote memories can be connected to the device executing the real-time production optimization method of the cracking furnace through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and their combinations.

[0133] The input device 503 can receive input user clicks and generate signal inputs related to user settings and function controls of the real-time production optimization method of the cracking furnace. The display device 504 may include a display screen and other display devices.

[0134] When the one or more modules are stored in the memory 502 and run by the one or more processors 501, the real-time production optimization method of the cracking furnace in any of the above method embodiments is executed.

[0135] In the present invention, key data is obtained by processing the historical operation data of the cracking furnace, then a target cracking depth model is established by analyzing the key data, and then multi-objective optimization is performed on the model to obtain a production objective function. The production of the cracking furnace is optimized in real time according to the production objective function and the collected real-time operation data. Thereby, it realizes improving the production benefit of the cracking furnace, monitoring the operation status of the cracking furnace in real time, and predicting operation failures in advance.

[0136] An embodiment of the present invention provides a storage medium that stores computer instructions. When a computer executes the computer instructions, it is used to perform all steps of the cracking furnace real-time production optimization method described above.

[0137] The above embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A real-time production optimization method for a cracking furnace, characterized in that, It includes: Collect and preprocess the historical operation data of the cracking furnace to obtain key data; Establish a target cracking depth model based on the analysis of the key data; Optimize the key data according to the target cracking depth model to obtain a production objective function; Obtain the real-time operation data of the cracking furnace; According to the production objective function and the real-time operation data, optimize the production of the cracking furnace in real time.

2. The real-time production optimization method for a cracking furnace according to claim 1, characterized in that, The collecting and preprocessing the historical operation data of the cracking furnace to obtain key data includes: Collect and sort out the historical operation data, and use the data smoothing method to clean the historical operation data to obtain the cleaned historical operation data; Perform correlation analysis on the cleaned historical operation data to obtain key data.

3. The real-time production optimization method for a cracking furnace according to claim 1, characterized in that, The establishing a target cracking depth model according to the analysis of the key data includes: Use the neural network algorithm to model the key data to obtain an initial cracking depth model; Obtain the model cracking depth and the actual cracking depth of the initial cracking depth model; Analyze and obtain the cracking depth deviation according to the model cracking depth and the actual cracking depth; According to the cracking depth deviation, iteratively optimize the initial cracking depth model to obtain the target cracking depth model.

4. The real-time production optimization method for a cracking furnace according to claim 3, characterized in that, The key data includes production operation data, production revenue data, and production consumption data. The production objective function includes a production revenue objective function and a production consumption objective function. The optimizing the key data according to the target cracking depth model to obtain a production objective function includes: According to the production operation data and the target cracking depth model, analyze and optimize the production revenue data to obtain the production revenue objective function; According to the production operation data and the production consumption data, analyze and optimize to obtain the production consumption objective function.

5. The real-time production optimization method for a cracking furnace according to claim 1, characterized in that, The optimizing the production of the cracking furnace in real time according to the production objective function and the real-time operation data includes: Speculate the production operation data according to the preset production target and the production objective function; Determine the operation state of the cracking furnace according to the real-time operation data; Determine whether there are operation hazards of the cracking furnace according to the real-time operation data.

6. The real-time production optimization method for a cracking furnace according to claim 5, characterized in that, The speculating the production operation data according to the preset production target and the production objective function includes: Input the production target as the dependent variable into the production objective function to speculate the production operation data.

7. The real-time production optimization method for a cracking furnace according to claim 5, characterized in that, The determining the operation state of the cracking furnace according to the real-time operation data includes: Calculate the thermal efficiency through the real-time operation data; Judge whether the thermal efficiency is less than the preset thermal efficiency value. If the thermal efficiency is less than the preset thermal efficiency value, prompt that the cracking furnace needs to be overhauled. If the thermal efficiency is greater than or equal to the preset thermal efficiency value, prompt that the operation state is normal.

8. The real-time production optimization method for a cracking furnace according to claim 5, characterized in that, The determining whether there are operation hazards of the cracking furnace according to the real-time operation data includes: Set the standard range of the key data; If the real-time operation data is not within the standard range, issue a warning of the operation hazard.

9. An electronic device, characterized in that, It includes: At least one processor; And, A memory communicatively connected to at least one of the processors; wherein, The memory stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors to enable at least one of the processors to execute the cracking furnace real-time production optimization method according to any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium stores computer instructions which, when executed by a computer, are used to execute all steps of the cracking furnace real-time production optimization method according to any one of claims 1 to 8.

Images