Best Practice Data-Driven Modeling and Optimization Approach for Heat Dispersion Systems in Papermaking Processes

Abstract

The present invention discloses a best practice data-driven modeling optimization method for the heat dispersion system of the papermaking process. Based on a large amount of historical operation data of the papermaking heat dispersion process collected, the energy cost per unit product is calculated to establish a best practice database with cost as the target. Outliers are removed from the database, and the most appropriate combination of processing parameters is suggested for difficult-to-decided processing parameter selection problems through automated modeling and intelligent optimization methods, which can reduce processing costs, energy consumption, etc. while improving processing quality and stabilizing industrial control.

Description

Technical Field

[0001] The present invention relates to the field of industrial process optimization, and in particular to a best practice data-driven modeling optimization method for a thermal dispersion system in a papermaking process. Background Art

[0002] A lot of impurities will be caught in the waste paper recycling process. There are a lot of substances that adhere to paper in the impurities, which are called stickies, mainly hot melt glue, adhesives, asphalt, grease, etc. The main function of thermal dispersion in the papermaking process is to refine and disperse the impurity particles so that they do not affect the subsequent production and product quality. There are problems such as extensive parameter adjustment management, no standards, and very serious energy waste in the current thermal dispersion process. This model is based on the parameters of samples with qualified quality indicators in the historical database, so the lowest cost samples in the historical data can be directly selected. The model has a small amount of calculation, fast calculation, and can be adjusted in real time. The parameter adjustment range is small, which can not only ensure the stable production, but also save costs and reduce energy consumption. Summary of the invention

[0003] The best practice data-driven modeling optimization method for the heat dispersion system of the papermaking process includes the following steps:

[0004] S1. Obtain the operating parameters of the papermaking heat dispersion process, process the abnormal values, and then establish the optimal operating parameter database according to the production cost;

[0005] S2. Establish operating parameter constraints and quality index constraints based on the principle of stable production of the heat dispersion process;

[0006] S3, establishing a heat dispersion cost function per liter of fiber for thermal and electrical coupling energy consumption during heat dispersion;

[0007] S4. Establish an iterative model for updating the optimal operating parameter database of heat dispersion. If it is not possible to find a better operating parameter for the current operating condition within the constraints, add the real-time operating condition to the optimal parameter database;

[0008] S5. Based on the constraints of stable operating parameters in the heat dispersion process and the goal of reducing production costs, a heat dispersion parameter optimization model is established to optimize the real-time production parameters and obtain the best practice operating parameters. If there are no recommended optimization parameters, the current working condition is returned to the updated iterative model.

[0009] Optimally, the specific steps of step S1 are as follows:

[0010] S11. After obtaining the historical data of thermal dispersion, first remove the data that does not meet the requirements of beating degree to ensure that the data in the optimal parameter database meets the normal production requirements.

[0011] S12. Process the abnormal values ​​of the thermal dispersion historical data, including data with a beating degree that does not meet the standard, abnormal operation data, and sensor error data. Use the box plot method to delete the abnormal operation data and sensor error data.

[0012] The process of the box plot method is to calculate the number of the lower quartile Q1 as position = (n+1) / 4 and the number of the upper quartile Q3 as position = 3(n+1) / 4, then the upper limit of the normal value is:

[0013] Q max =Q3+1.5(Q3-Q1)

[0014] The lower limit of normal value is:

[0015] Q min =Q3-1.5(Q3-Q1)

[0016] Optimally, the specific steps of step S2 are as follows:

[0017] S21. In order to ensure stable production, the adjustment of the production process must be kept within a certain range, and the operating parameter constraints for stable production of the thermal dispersion process include: back pressure power constraint, thermal dispersion power constraint and slurry temperature constraint. After the thermal dispersion process, it is necessary to ensure that the slurry beating degree meets the requirements, so the constraints also include the quality indicator beating degree constraint.

[0018] The back pressure power constraint is:

[0019] μ b，min F b，a <F b <μ b，max F b，a

[0020] Among them, F b，a Indicates the back pressure power in actual operation, F b Represents the optimized back pressure power, μ b，min , μ b，max They represent the minimum and maximum back pressure constraint coefficients under stable production conditions respectively.

[0021] The heat dissipation power constraint is:

[0022] μ e，min F e，a <F e <μ e，max F e，a

[0023] Among them, F e，a Indicates the actual heat dissipation power in operation, F e represents the optimized heat dissipation power, μ e，min , μ e，maxThey represent the minimum and maximum heat dispersion constraint coefficients under stable production conditions respectively.

[0024] The slurry temperature constraint is:

[0025] μ t，min T f，a <T f <μ t，max T f，a

[0026] Among them, T f，a Indicates the actual slurry temperature during operation, T f represents the optimized slurry temperature, μ t，min , μ t，max They respectively represent the minimum and maximum slurry temperature constraint coefficients under stable production conditions.

[0027] Hereμ b，min , μ b，max , μ e，min , μ e，max , μ t，min , μ t，max They are collectively called constraint coefficients. The constraint coefficients of the optimization process in different situations can be changed according to the tendency of the optimization target. Taking environmental factors into consideration, it is more inclined to reduce the use of steam, which can reduce the slurry temperature constraint coefficient; simply considering lower costs can expand the optimization range, that is, reducing the lower limit and increasing the upper limit; considering the stability of production, the optimization range can be reduced.

[0028] Optimally, the specific steps of step S3 are as follows:

[0029] S31. The heat dispersion process involves the coupled use of two energy sources, namely electricity and steam. In addition, due to factors such as production adjustment and raw material adjustment, the flow rate and fiber concentration of the slurry are different at different times. Generally, statistical energy consumption by time cannot well reflect the optimization effect of operating parameter changes on production costs. Therefore, the present invention uses the energy cost consumed by heat dispersion of 1 liter of absolute dry fiber as the objective function. The objective function of the heat and electricity coupled energy optimization model of the heat dispersion process is:

[0030]

[0031] Optimally, the specific steps of step S5 are as follows:

[0032] S51. Read the back pressure power, heat dispersion power, and slurry temperature of the current real-time working condition, and calculate the back pressure power constraint, heat dispersion power constraint, and slurry temperature constraint.

[0033] S52. Read the back pressure power, heat dispersion power, and steam flow rate of the current real-time working condition, and calculate the cost required for heat dispersion of 1 liter of absolute dry fiber under the current working condition.

[0034] S53. Search the optimal parameter database for historical data that meets the back pressure power constraint, heat dispersion power constraint, and slurry temperature constraint, and compare the historical data with the cost of the current real-time operating condition. If the cost is found to be less than the current operating condition, add this data to the optimized data set.

[0035] S54. Repeat S53 until the entire optimal parameter database is traversed, and output the back pressure power, heat dispersion power and slurry temperature of the lowest cost historical data in the optimized data set.

[0036] S55. If the optimized data set is 0 after the cycle ends, the current real-time working condition is added to the optimal parameter database.

[0037] The technical effects of the present invention are as follows:

[0038] 1. Analyze and clean up the historical data of the heat dispersion process, remove the data of non-normal operation status, remove the sensor error data, and ensure that the data in the optimal parameter database are all normal operation data.

[0039] 2. The constraint coefficient is set in the present invention to ensure the stability of production to the greatest extent. At the same time, different optimization effects can be achieved by changing the constraint coefficient according to different optimization target tendencies.

[0040] 3. Based on the coupled use of thermal energy and electrical energy in the heat dispersion process, the present invention adopts the energy cost consumed by unit absolute dry fiber of heat dispersion as the objective function, which reduces the impact of raw material adjustment and production adjustment on energy calculation and can better show the cost reduction effect after optimization.

[0041] 4. The present invention has an optimal parameter database iteration mechanism. When more optimal operating parameters are found, they will be automatically added to the optimal parameter database, which can ensure that the optimization effect is continuously improved within the normal operating range. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic diagram of the present invention;

[0043] Figure 2a is the fiber flow rate history data before the abnormal value is processed in the embodiment;

[0044] Figure 2b is the fiber flow historical data after the abnormal value is processed in the embodiment;

[0045] Figure 3 This is a back pressure power optimization effect diagram after the real-time operating parameters of Example 19 are optimized;

[0046] Figure 4 This is a diagram showing the optimization effect of heat dispersion power after the optimization of the real-time operating parameters in Example 19;

[0047] Figure 5 This is a diagram showing the slurry temperature optimization effect after the real-time operating parameters of Example 19 are optimized;

[0048] Figure 6 This is a cost optimization effect diagram after the real-time operating parameters of Example 19 are optimized. DETAILED DESCRIPTION

[0049] The present invention will be further described below in conjunction with the embodiments of the present invention, and the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments completed by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] like Figure 1 As shown, the modeling steps are:

[0051] 1. Collect the historical data of thermal dispersion operation within a certain period of time, process the outliers in the data, use the box plot to eliminate the outliers, and calculate the lower quartile Q1 as the number of position = (n+1) / 4 and the upper quartile Q3 as the number of position = 3(n+1) / 4. Then the upper limit of the normal value Q max for:

[0052] Q max =Q3+1.5(Q3-Q1)

[0053] Lower limit of normal value Q min for:

[0054] Q min =Q3-1.5(Q3-Q1)

[0055] Eliminate the abnormal values ​​outside the upper and lower limits, and the fiber flow before and after treatment is as follows Figure 1 As shown in the figure, the processed data is obviously smoother, with fewer extreme data that do not conform to normal operation.

[0056] 2. Classify the data based on fiber flow rate and use clustering algorithm to divide different working conditions.

[0057] 3. Determine the cost function, where the amount of electricity and steam consumed to heat disperse 1 liter of absolute dry fiber is multiplied by the unit price of electricity and the unit price of steam to obtain the cost function for heat dispersing 1 liter of absolute dry fiber.

[0058] 4. Determine the constraint coefficient, where μ b，min , μ b，max , μ e，min , μ e，max , μ t，min , μ t，maxThe values ​​are 0.98, 1.02, 0.99, 1.01, 0.95, and 1.05 respectively. The constraint coefficient can be adjusted according to actual needs. Considering environmental protection, the lower limit constraint of steam flow can be increased to minimize the use of steam. Improving production stability can make all coefficients as close to 1 as possible.

[0059] 5. With the goal of minimizing the cost function, start optimizing under the constraints, and return the optimal back pressure power, heat dispersion power and fiber flow rate after the optimization is completed.

[0060] 6. If there are no better parameters, add the actual operating conditions of this group to the optimal operating parameter database and iteratively update the database.

[0061] This embodiment tests and analyzes 20 groups of real-time operating parameters, of which 19 groups recommend optimized parameters, and 1 group of operating conditions is judged as a more optimal operating condition and added to the optimal operating parameter database. Figure 2a , Figure 2b , Figure 3 , Figure 4 , which are 19 groups of effect diagrams of back pressure power, heat dispersion power and slurry temperature. Figure 5 Cost-optimized renderings.

Claims

1. A best practice data-driven modeling optimization method for heat dispersion system in papermaking process, characterized by: The following steps are involved: S1. Obtain the operating parameters of the papermaking heat dispersion process, process the abnormal values ​​of the heat dispersion historical data, and then establish the optimal operating parameter database according to the production cost; S2. Establish operating parameter constraints and quality index constraints based on the principle of stable production of the heat dispersion process; In order to ensure stable production, the adjustment of the production process must be kept within a certain range. The operating parameter constraints for stable production in the thermal dispersion process include: back pressure power constraint, thermal dispersion power constraint and slurry temperature constraint. After the thermal dispersion process, it is necessary to ensure that the slurry beating degree meets the requirements, so the constraints also include the quality indicator beating degree constraint. The back pressure power constraint is: m b，min F b，a <F b <m b，max F b，a Among them, F b，a Indicates the back pressure power in actual operation, F b Represents the optimized back pressure power, μ b，min , μ b，max They represent the minimum and maximum back pressure constraint coefficients under stable production conditions respectively; The heat dissipation power constraint is: m e，min F e，a <F e <m e，max F e,a Among them, F e,a Indicates the actual heat dissipation power in operation, F e represents the optimized heat dissipation power, μ e，min , μ e，max They represent the minimum and maximum heat dispersion constraint coefficients under stable production conditions; The slurry temperature constraint is: μ t，min T f，a <T f <μ t，max T f，a Among them, T f，a Indicates the actual slurry temperature during operation, T f represents the optimized slurry temperature, μ t，min , μ t，max They represent the minimum and maximum slurry temperature constraint coefficients under stable production conditions respectively; Hereμ b，min , μ b，max , μ e，min , μ e，max , μ t，min , μ t，max They are collectively called constraint coefficients, and the constraint coefficients of the optimization process in different situations change according to the tendency of the optimization goal; S3, establishing a heat dispersion cost function per liter of fiber for thermal and electrical coupling energy consumption during heat dispersion; where a p represents the unit price of electricity, b p Indicates the unit price of steam, F s Indicates the steam flow rate, Q f is the fiber flow rate, l c Indicates fiber concentration; S4. Establish an update and iteration model for the optimal operating parameter database of heat dispersion. If a better operating parameter of the current working condition cannot be found within the constraints, the operating parameter of the current working condition is added to the optimal parameter database; S5. According to the constraints of the operating parameters for stable production of the heat dispersion process and the goal of reducing production costs, a heat dispersion parameter optimization model is established to optimize the real-time production parameters and obtain the best practice operating parameters; The best practice operating parameters include: recommended optimal back pressure power, optimal heat dispersion power and optimal steam temperature; if no optimization parameters are recommended, the current operating condition is saved as the optimal operating condition parameter; If there are no recommended optimization parameters, the current working condition will be returned and the iterative model will be updated.

2. The best practice data-driven modeling optimization method for heat dispersion system in papermaking process according to claim 1 is characterized in that: In S1, the abnormal values ​​of the thermal dispersion historical data, including data with a beating degree that does not meet the standard, abnormal operation data and sensor error data, are processed; Data that do not meet the standard are directly deleted, and then the abnormal operation data and sensor error data are deleted using the box plot method; The process of the box plot method is to calculate the number of the lower quartile Q1 as position = (n+1) / 4 and the number of the upper quartile Q3 as position = 3(n+1) / 4, then the upper limit of the normal value is: Q max =Q3+1.5(Q3-Q1) The lower limit of normal value is: Q min = Q3-1.5(Q3-Q1).

Images