Method for optimizing fresh gas consumption per ton of ammonia in ammonia synthesis workshop based on industrial big data modeling

Abstract

The invention provides a method for optimizing fresh gas consumption per ton of ammonia in a synthesis ammonia workshop based on industrial big data modeling. The method comprises the following steps: finding key influence parameters and variables influencing fresh gas consumption per ton of ammonia in natural gas synthesis ammonia (namely a target function); key influence parameters and variables of key operation limit values (namely constraint conditions) influencing natural gas ammonia synthesis are found; performing data preprocessing on historical production data of the key parameters and variables to obtain healthy data for model training; training a ton ammonia fresh gas consumption model; training a key operation limit value model; and solving the nonlinear optimization problem of fresh gas consumption per ton of ammonia by combining the objective function and the constraint condition. According to the method, the natural gas synthesis ammonia production optimization model is established, the set value of each key control parameter is searched by using the algorithm to guide production, the fresh gas consumption per ton of ammonia is minimum, and the dependence of a production line on the experience of workers is eliminated.

Description

Technical Field

[0001] The present invention belongs to the field of optimization control of natural gas synthesis ammonia production process, and specifically relates to a method for optimizing the consumption of fresh gas per ton of synthetic ammonia based on industrial big data modeling. Background Art

[0002] The production of natural gas-synthesized ammonia uses natural gas, air and steam as basic raw materials. It goes through the steps of desulfurization, conversion, transformation, decarbonization, purification, compression and synthesis (freezing) to ultimately produce a product with an ammonia content of more than 99.8%. Natural gas-synthesized ammonia is a typical long-process chemical process.

[0003] The existing synthetic ammonia control method is that, under the premise of meeting the safety of chemical production, engineers write appropriate and reasonable operating parameters on process cards and specify the control range of key operating parameters (for example, the first outlet temperature, converter outlet temperature, synthesis tower hot spot temperature, hydrogen-nitrogen ratio, water-carbon ratio, and joint compressor outlet pressure, etc.). When the production conditions change, the operators make adjustments based on the actual operation conditions to ensure that the parameters are within the control range, so as to achieve the goals of safe, stable, long-term, full, and excellent chemical production.

[0004] At present, the method of adjusting production based only on the range of key control parameters often encounters the following problems in actual operation:

[0005] (1) The composition of the raw natural gas will change during the production process: When the composition of the gas source changes, workers cannot obtain the current composition information of the raw natural gas. Due to the strong coupling relationship between the key parameters, workers need to adjust the production system back to a steady state based on past experience. The length of time it takes is related to the workers' experience and concentration.

[0006] (2) Natural gas synthesis of ammonia is a dynamic and continuous process, which is affected by many uncertain factors. Operators can only adjust parameters based on experience. Therefore, the production cost of synthetic ammonia is closely related to the workers' operating skills, experience, and concentration. In addition, the workers' operation is based on years of accumulated experience, which is the feasible value of the current operation rather than the optimal value. There is room for improvement in its production efficiency. Summary of the invention

[0007] In view of the above technical problems, the present invention provides a method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia workshop based on industrial big data modeling. Based on the production process of natural gas synthetic ammonia, the mechanism analysis methods of mass transfer, heat transfer, momentum transfer, and reaction engineering are adopted to select the key control parameter training model of the natural gas synthetic ammonia process to achieve accurate prediction of ammonia production. Based on the trained prediction model and real-time production data, the parameters of each section of synthetic ammonia are jointly optimized to achieve the economic goal of minimizing the fresh gas consumption per ton of ammonia while ensuring production safety.

[0008] The method first pre-processes a large amount of industrial historical data and removes abnormal samples in the historical data. Then, it selects the key influencing parameters of the natural gas synthesis ammonia process to establish a mathematical model, and uses algorithms to find the optimal solution of the variables to guide production, ensure the optimal fresh gas consumption per ton of ammonia, and get rid of the reliance on workers' experience.

[0009] The specific technical solutions are:

[0010] The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia workshop based on industrial big data modeling includes the following steps:

[0011] Step S100: finding the key influencing parameters and variables that affect the fresh gas consumption per ton of ammonia synthesized from natural gas (i.e., the objective function);

[0012] These key variables and parameters include:

[0013] y1 (liquid ammonia flow): fresh gas flow x1, synthesis tower lower layer temperature ya3, ice machine liquid ammonia inlet ammonia cooling temperature x11, 2# ammonia cooler shell side outlet flow x13, methanation gas methane content x2, fresh gas hydrogen content x3, joint compressor outlet pressure x5, purge gas flow x7, synthesis tower upper layer temperature ya1, synthesis tower middle layer temperature ya2.

[0014] Step S200: finding key influencing parameters and variables that affect key operating limits (i.e., constraints) of natural gas synthesis of ammonia;

[0015] These key variables and parameters include:

[0016] Conversion section:

[0017] y1 (methane output): daily steam flow rate of the first converter x1, water-carbon ratio x2, negative pressure of the first converter x3, pressure entering the ammonia converter x4, temperature entering the first converter x5, daily combustion flow rate x6, combustion flow rate of the membrane first converter x7, pressure difference between the inlet and outlet of the first converter x8, temperature difference inside the first converter x9.

[0018] y2 (first outlet temperature): first converter daily steam flow rate x1, water-carbon ratio x2, first converter negative pressure x3, first converter inlet pressure x4, first converter inlet temperature x5, fuel flow rate x6, membrane first converter combustion flow rate x7, first converter inlet and outlet pressure difference x8, first converter internal temperature difference x9.

[0019] y3 (second stage furnace temperature): first outlet methane ya1, converter natural steam flow x1, first outlet temperature ya2, second converter reaction pressure x13, natural steam flow into converter x14, converter outlet temperature x15, air flow x16, air temperature x17, second converter pressure difference x18, second converter cooling water enthalpy change x19.

[0020] y4 (secondary methane output): primary methane output ya1, natural steam flow rate of converter x1, primary output temperature ya2, secondary converter reaction pressure x13, natural steam flow rate into converter x14, converter outlet temperature x15, air flow rate x16, air temperature x17, secondary converter pressure difference x18.

[0021] y5 (converter outlet temperature): natural steam flow rate of the first converter x1, natural steam flow rate into the converter x14, second-stage furnace furnace temperature ya3, natural steam mixed gas temperature x11, converter cooling water enthalpy change x12, second outlet methane ya4, second outlet pressure x20, converter cold side temperature difference x21.

[0022] Synthesis section:

[0023] y2 (temperature of the upper layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, combined compressor outlet pressure x5, synthesis gas flow x6.

[0024] y3 (temperature of the middle layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, joint compressor outlet pressure x5, synthesis gas flow x6.

[0025] y4 (temperature of the lower layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, joint compressor outlet pressure x5, synthesis gas flow x6.

[0026] Step S300: preprocess the historical production data of the above key parameters and variables to obtain healthy data for model training;

[0027] Step S310: Due to interference from the production environment at the synthetic ammonia production site or a failure at a measuring point at a certain moment, the collected data show that the limit value is exceeded, and these abnormal values ​​are eliminated;

[0028] Step S320: For data points with frequent production fluctuations, data processing needs to be performed through data filtering;

[0029] After preprocessing, historical production data is used to train the objective function and constraint model.

[0030] Step S400: training a fresh gas consumption model per ton of ammonia;

[0031] Step S410:

[0032] The variables are "normalized" and the processing formula is:

[0033]

[0034] xoN is a dimensioned prediction variable, Max{xoN} is the maximum value of the xoN array, and Min{xoN} is the minimum value of the xoN array. These two values ​​are determined based on the trained historical data set.

[0035] Step S420:

[0036] The model equation for liquid ammonia production is established as follows:

[0037] y1=e=f(x1,ya3,x11,x13,x2,x3,x5,rx7,ya1,ya2)+y1_err;

[0038] Among them, xn is a parameter, rxn is a variable, and yn_err is calculated in real time according to the following formula.

[0039] Step S500: training a key operation limit model;

[0040] Step S510:

[0041] The variables are "normalized" and the processing formula is:

[0042]

[0043] xoN is a dimensioned prediction variable, Max{xoN} is the maximum value of the xoN array, and Min{xoN} is the minimum value of the xoN array. These two values ​​are determined based on the trained historical data set.

[0044] Step S520:

[0045] A model equation for methane is established:

[0046] y1=e=f(rx1, rx2, rx3, x4, x5, rx6, x7, x8, x9)+y1_err;

[0047] Step S530:

[0048] A temperature model equation is established as follows:

[0049] y2=e=f(rx1, rx2, rx3, x4, x5, rx6, x7, x8, x9)+y2_err;

[0050] Step S540:

[0051] The model equation of the second-stage furnace temperature is established as follows:

[0052] y3=e=f(ya1, rx1, ya2, x13, rx14, x15, x16, x17, x18, x19)+y3_err;

[0053] Step S550:

[0054] The model equation for di-methane is established as:

[0055] y4=e=f(ya1, rx1, ya2, x13, rx14, x15, x16, x17, x18, x10)+y4_err;

[0056] Step S560:

[0057] The model equation for converter outlet temperature is established as follows:

[0058] y5=e=f(rx1, rx14, ya3, x11, x12, ya4, z20, x21)+y5_err;

[0059] Step S570:

[0060] The model equation for the temperature of the upper layer of the synthesis tower is established as follows:

[0061] y2=e==f(x14,x15,x16,rx4,x5,x6)+y2_err;

[0062] Step S580:

[0063] The model equation for the temperature in the middle layer of the synthesis tower is established as follows:

[0064] y3=e=f(x14,x15,x16,rx4,x5,x6)+y3_err;

[0065] Step S590:

[0066] The model equation for the temperature of the lower layer of the synthesis tower is established as follows:

[0067] y4=e=f(x14,x15,x16,rx4,x5,x6)+y4_err;

[0068] In steps S520 to S590,

[0069] Among them, xn is a parameter, rxn is a variable, and yn_err is calculated in real time according to the following formula.

[0070] Step S600: Combine the objective function and the constraints to solve the nonlinear optimization problem of fresh gas consumption per ton of ammonia.

[0071] Combine the equations of liquid ammonia flow rate, first outlet methane, first outlet temperature, second stage furnace temperature, second outlet methane, converter outlet temperature, synthesis tower upper layer temperature, synthesis tower middle layer temperature, synthesis tower lower layer temperature, with the Max liquid ammonia flow rate as the goal, use mathematical programming technology, and adopt the classic continuous convex approximation algorithm to efficiently solve each variable.

[0072] The technical effects of the present invention are as follows:

[0073] (1) Based on the principles of mass transfer, heat transfer, momentum transfer and reaction engineering of natural gas synthesis of ammonia, determine the key production parameters and variables that affect the fresh gas consumption per ton of ammonia, the first outlet methane, the first outlet temperature, the furnace temperature of the second-stage reformer, the outlet temperature of the converter, the second outlet methane, the upper temperature of the synthesis tower, the middle temperature of the synthesis tower, and the lower temperature of the synthesis tower.

[0074] (2) When there is much interference in the production data of synthetic ammonia production or a measurement point fails at a certain moment, the collected data will exceed the limit. These abnormal values ​​need to be automatically eliminated. Data filtering should be performed on data points with frequent production fluctuations to ensure that the optimization recommendation is stable and reliable.

[0075] (3) Use the model to jointly solve key parameters, find feasible operating conditions under production safety conditions, and finally formulate a joint adjustment strategy for various production parameters with the optimal production cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a schematic diagram of the method of the present invention;

[0077] Figure 2 is the prediction result of the objective function;

[0078] Figure 3 shows the prediction results of each constraint condition. DETAILED DESCRIPTION

[0079] The specific technical solution of the present invention is explained in conjunction with embodiments.

[0080] Taking a natural gas ammonia synthesis plant as an example, the key control indicators of the plant are collected as constraints. The key control indicators of the conversion section are shown in Table 1-1, and the key control indicators of the synthesis section are shown in Table 1-2.

[0081] Constraint 1:

[0082] Table 1-1 Key index control range of the conversion section of a natural gas synthetic ammonia plant

[0083] response Indicator name Lower limit Upper limit unit y1 Temperature 690 760 ℃ y2 Methane 0 13 ％ y3 Second stage furnace temperature 880 984 ℃ y4 Dimethazone 0 0.6 ％ y5 Converter outlet temperature 720 760 ℃

[0084] Table 1-2 Key index control range of the synthesis section of a natural gas synthetic ammonia plant

[0085]

[0086]

[0087] Constraint 2: Changes in all control parameters are within the operating range allowed by the process.

[0088] Constraint three: All control parameters are adjusted within + / -2% of the current operating level to ensure the continuity and stability of production adjustments.

[0089] Based on the above conditions, use Figure 1 The method shown is deployed and implemented.

[0090] Step S100: finding the key influencing parameters and variables that affect the fresh gas consumption per ton of ammonia synthesized from natural gas (i.e., the objective function);

[0091] Step S200: finding key influencing parameters and variables that affect key operating limits (i.e., constraints) of natural gas synthesis of ammonia;

[0092] Step S300: Take the factory's production data for 90 consecutive days with a time interval of 1 minute, and obtain a total of 129,600 sets of production data. Remove the abnormal values ​​in the production data, perform an average filtering on the data for 15 minutes, and obtain more than 8,500 sets of valid data.

[0093] Step S400: training a fresh gas consumption model per ton of ammonia;

[0094] Step S500: training a key operation limit model;

[0095] Step S600: Combine the objective function and the constraints to solve the nonlinear optimization problem of fresh gas consumption per ton of ammonia.

[0096] Based on the results of the factory control experiment, the optimized fresh gas volume per ton of ammonia can achieve an improvement of more than 1%.

Claims

1. A method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant based on industrial big data modeling, characterized in that: The following steps are involved: Step S100: finding the key influencing parameters and variables that affect the fresh gas consumption per ton of ammonia synthesized from natural gas (i.e., the objective function); Step S200: finding key influencing parameters and variables that affect key operating limits (i.e., constraints) of natural gas synthesis of ammonia; Step S300: preprocess the historical production data of the above key parameters and variables to obtain healthy data for model training; Step S400: training a fresh gas consumption model per ton of ammonia; Step S500: training a key operation limit model; Step S600: Combine the objective function and the constraints to solve the nonlinear optimization problem of fresh gas consumption per ton of ammonia.

2. The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant based on industrial big data modeling according to claim 1, characterized in that: In step S100, the key influencing parameters and variables of fresh gas consumption per ton of ammonia include: y1 (liquid ammonia flow): fresh gas flow x1, synthesis tower lower layer temperature ya3, ice machine liquid ammonia inlet ammonia cooling temperature x11, 2# ammonia cooler shell side outlet flow x13, methanation gas methane content x2, fresh gas hydrogen content x3, joint compressor outlet pressure x5, purge gas flow x7, synthesis tower upper layer temperature ya1, synthesis tower middle layer temperature ya2.

3. The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant based on industrial big data modeling according to claim 1, characterized in that: Step S200, find the key operating limits (i.e., constraints) and their corresponding key influencing parameters and variables, which are: Conversion section: y1 (methane output): daily steam flow rate of the first converter x1, water-carbon ratio x2, negative pressure of the first converter x3, pressure of the first converter entering the ammonia converter x4, temperature of the first converter entering the first converter x5, daily combustion flow x6, combustion flow rate of the first converter x7, pressure difference of the first converter inlet and outlet x8, temperature difference of the first converter x9; y2 (first outlet temperature): first converter steam flow x1, water-carbon ratio x2, first converter negative pressure x3, first converter pressure x4, first converter temperature x5, first converter flow x6, first converter combustion flow x7, first converter inlet and outlet pressure difference x8, first converter internal temperature difference x9; y3 (second stage furnace temperature): first outlet methane ya1, converter natural steam flow x1, first outlet temperature ya2, second converter reaction pressure x13, natural steam flow into converter x14, converter outlet temperature x15, air flow x16, air temperature x17, second converter pressure difference x18, second converter cooling water enthalpy change x19; y4 (methane at the second outlet): methane at the first outlet ya1, natural steam flow rate of converter x1, temperature at the first outlet ya2, reaction pressure of the second converter x13, natural steam flow rate into the converter x14, converter outlet temperature x15, air flow rate x16, air temperature x17, pressure difference of the second converter x18; y5 (converter outlet temperature): natural steam flow rate of the first converter x1, natural steam flow rate into the converter x14, second stage furnace hearth temperature ya3, natural steam mixed gas temperature x11, converter cooling water enthalpy change x12, second outlet methane ya4, second outlet pressure x20, converter cold side temperature difference x21; Synthesis section: y2 (temperature of the upper layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, joint compressor outlet pressure x5, synthesis gas flow x6; y3 (temperature of the middle layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, joint compressor outlet pressure x5, synthesis gas flow x6; y4 (temperature of the lower layer of the synthesis tower): 1# main line flow x14, 2# main line flow x15, circulating gas methane content x16, hydrogen-nitrogen ratio x4, joint compressor outlet pressure x5, synthesis gas flow x6.

4. The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant by industrial big data modeling according to claim 1 is characterized in that: The modeling process in step S400 is: Step S410: The variables are "normalized" and the processing formula is: xoN is a dimensioned prediction variable, Max{xoN} is the maximum value of the xoN array, and Min{xoN} is the minimum value of the xoN array. These two values ​​are determined based on the historical data set for training; Step S420: The model equation for liquid ammonia production is established as follows: y1=e=f(x1,ya3,x11,x13,x2,x3,x5,rx7,ya1,ya2)+y1_err; Among them, xn is a parameter, rxn is a variable, and yn_err is calculated in real time according to the following formula: in_err=in 实际值 -in 训练值 Among them, 实际值 is the average value of the actual value of yn during period t, yn 训练值 is the average value of yn training value in period t, and the value range of t is: 1-120min.

5. The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant by industrial big data modeling according to claim 1 is characterized in that: The modeling process in step S500 is: Step S510: The variables are "normalized" and the processing formula is: xoN is a dimensioned prediction variable, Max{xoN} is the maximum value of the xoN array, and Min{xoN} is the minimum value of the xoN array. These two values ​​are determined based on the historical data set for training; Step S520: A model equation for methane is established: y1=e=f(rx1, rx2, rx3, x4, x5, rx6, x7, x8, x9)+y1_err; Step S530: A temperature model equation is established as follows: y2=e=f(rx1, rx2, rx3, x4, x5, rx6, x7, x8, x9)+y2_err; Step S540: The model equation of the second-stage furnace temperature is established as follows: y3=e=f(ya1, rx1, ya2, x13, rx14, x15, x16, x17, x18, x19)+y3_err; Step S550: The model equation for di-methane is established as: y4=e=f(ya1, rx1, ya2, x13, rx14, x15, x16, x17, x18, x10)+y4_err; Step S560: The model equation for converter outlet temperature is established as follows: y5=e=f(rx1, rx14, ya3, x11, x12, ya4, x20, x21)+y5_err; Step S570: The model equation for the temperature of the upper layer of the synthesis tower is established as follows: y2=e=f(x14,x15,x16,rx4,x5,x6)+y2_err; Step S580: The model equation for the temperature in the middle layer of the synthesis tower is established as follows: y3=e=f(x14,x15,x16,rx4,x5,x6)+y3_err; Step S590: The model equation for the temperature of the lower layer of the synthesis tower is established as follows: y4=e=f(x14,x15,x16,rx4,x5,x6)+y4_err; In steps S520 to S590, Among them, xn is a parameter, rxn is a variable, and yn_err is calculated in real time according to the following formula. in_err=in 实际值 -in 训练值 Among them, 实际值 is the average value of the actual value of yn during period t, yn 训练值 is the average value of yn training value in period t, and the value range of t is: 1-120min.

6. The method for optimizing the fresh gas consumption per ton of ammonia in a synthetic ammonia plant based on industrial big data modeling according to claim 1, characterized in that: In step S600, the following equations are combined: liquid ammonia flow rate, first outlet methane, first outlet temperature, second stage furnace temperature, second outlet methane, converter outlet temperature, synthesis tower upper layer temperature, synthesis tower middle layer temperature, synthesis tower lower layer temperature. With the goal of maximizing the liquid ammonia flow rate, mathematical programming technology is used to efficiently solve each variable using a classic continuous convex approximation algorithm.

Images