A multi-stage combustion injection system for a float glass melting furnace

Abstract

The present application relates to the technical field of float glass production, and proposes a multi-stage combustion injection system for a float glass melting furnace. The system comprises a multi-stage injection unit, a detection unit and a control unit. The multi-stage injection unit comprises a first-stage fuel injector, a second-stage combustion-supporting gas injector and a third-stage flame shaping injector. The control unit generates a heat flux density distribution and a temperature distribution in a calculation area of a free surface of glass liquid, calculates a co-location deviation between a heat flux concentration area and a temperature response area, and generates an injection correction amount according to the co-location deviation and a glass liquid surface temperature deviation to adjust the injection amount of each stage, so as to realize collaborative control of a flame heat input position, a lateral coverage range and a glass liquid temperature response area.

Description

Technical Field

[0001] This invention relates to the field of float glass production technology, and in particular to a multi-stage combustion injection system for float glass melting furnaces. Background Technology

[0002] The melting section of a float glass melting furnace requires a continuous input of heat to the free surface of the molten glass via a flame to complete the melting of the batch, the clarification of the molten glass, and the maintenance of the thermal regime. Existing furnace combustion control typically adjusts fuel flow rate, combustion gas flow rate, combustion space temperature, molten glass surface temperature, flue gas oxygen content, carbon monoxide content, or furnace pressure to maintain a certain overall temperature level. However, it is difficult to determine whether the flame heat input area and the molten glass temperature response area are spatially matched. When the flame heat release position is advanced or delayed along the length of the furnace, or when the heat flow concentration area is biased to one side along the width of the furnace, simply adjusting the fuel flow rate or total combustion gas flow rate is insufficient to separately correct the heat release position in the middle and later stages of the flame and the lateral coverage area of ​​the flame, easily causing mutual interference between adjustment actions. Therefore, a multi-stage combustion injection system is needed that can simultaneously obtain the heat flux density distribution and temperature distribution within the same calculated area of ​​the molten glass free surface, and, based on the co-location deviation between the heat flow concentration area and the temperature response area, perform graded corrections for primary fuel injection, secondary combustion gas injection, and tertiary flame shaping injection. Summary of the Invention

[0003] This invention addresses the problems of existing float glass melting furnace combustion control systems, such as the difficulty in determining whether the flame heat input area and the glass melt temperature response area are co-located, and the difficulty in separately correcting the longitudinal heat release position and lateral coverage of the flame. It proposes a multi-stage combustion injection system for float glass melting furnaces. This system is applied to the combustion station in the melting section of the float glass melting furnace and includes a multi-stage injection unit, a detection unit, and a control unit. The multi-stage injection unit includes a primary fuel injector for forming the flame root, a secondary combustion-supporting injector for supplementing the middle and rear sections of the flame, and a tertiary flame-shaping injector for adjusting the lateral coverage of the flame. The detection unit acquires the furnace operating status parameters. The control unit adjusts the furnace operating status parameters and the current injection status of each injector according to the current injection status of each stage. The injection parameters are set up by establishing a thermal coupling interface corresponding to the free surface of the molten glass between the combustion space sub-model and the molten glass sub-model. This generates the heat flux density distribution and temperature distribution within the same molten glass free surface calculation area. Based on the heat flux density distribution and temperature distribution, the longitudinal and lateral deviation values ​​between the heat flux concentration area and the temperature response area, as well as the molten glass surface temperature deviation, are determined. The control unit further generates injection correction values ​​at each level based on the longitudinal deviation values, lateral deviation values, and molten glass surface temperature deviation. Based on the injection correction values ​​at each level, the first-stage fuel injector, the second-stage combustion-supporting injector, and the third-stage flame-shaping injector are adjusted respectively to achieve graded and coordinated adjustment of the heat input intensity at the flame root, the heat release position in the middle and rear sections of the flame, and the lateral coverage range of the flame.

[0004] This invention proposes a multi-stage combustion injection system for float glass melting furnaces, applied to a combustion station in the melting section of the float glass melting furnace. The float glass melting furnace includes a combustion space and a free surface of molten glass. The system comprises: a multi-stage injection unit, including a primary fuel injector, a secondary combustion-supporting injector, and a tertiary flame-shaping injector; a detection unit for acquiring furnace operating status parameters; and a control unit, including a combustion-melting coupling prediction unit, a co-position deviation calculation unit, a predictive control optimization unit, and an execution allocation unit. The combustion-melting coupling prediction unit includes a combustion space sub-model, a molten glass sub-model, and a thermal coupling interface, which transmits the furnace operating status parameters and the current injection status of the primary fuel injector, secondary combustion-supporting injector, and tertiary flame-shaping injector. Using injection parameters as input, a thermal coupling interface corresponding to the free surface of the molten glass is set between the combustion space sub-model and the molten glass sub-model to generate heat flux density distribution and temperature distribution; a co-location deviation calculation unit calculates the co-location deviation based on the heat flux density distribution and temperature distribution; and determines the molten glass surface temperature deviation based on the difference between the temperature distribution and the preset target temperature distribution; a predictive control optimization unit includes an injection correction strategy model; the injection correction strategy model uses the co-location deviation and the molten glass surface temperature deviation as input to generate injection correction weights and further calculate the injection correction amount; and an execution allocation unit adjusts the fuel injection amount of the first-stage fuel injector, the second-stage combustion gas injection, and the third-stage flame shaping gas injection amount according to the injection correction amount.

[0005] Furthermore, the furnace operating parameters include the flow rate of each injection stage, the temperature of the combustion space, the surface temperature of the molten glass, the oxygen content of the flue gas, the carbon monoxide content of the flue gas, the furnace pressure, and the molten glass level.

[0006] Furthermore, the colocation deviation calculation unit determines the heat flux concentration area based on the heat flux density distribution, determines the temperature response area based on the temperature distribution, and calculates the colocation deviation between the heat flux concentration area and the temperature response area. The colocation deviation includes the longitudinal deviation value in the length direction of the melting furnace and the lateral deviation value in the width direction of the melting furnace. Based on the difference between the temperature distribution and the preset target temperature distribution, the surface temperature deviation of the molten glass is determined.

[0007] Furthermore, the predictive control optimization unit includes an injection correction strategy model; the injection correction strategy model includes a sparse simplex weight generation layer and a polyhedral constraint safety correction layer; the sparse simplex weight generation layer takes the longitudinal deviation value, the lateral deviation value, and the glass melt surface temperature deviation as inputs to generate the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection, and ensures that the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection satisfy the non-negative normalization constraint; the polyhedral constraint safety correction layer constructs an injection correction feasible region based on the furnace operating state parameters, and performs constraint correction on the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection within the injection correction feasible region to obtain the first-stage fuel injection correction weights, the second-stage combustion gas injection correction weights, and the third-stage shaping gas injection correction weights; and further calculates the first-stage fuel injection correction amount, the second-stage combustion gas injection correction amount, and the third-stage shaping gas injection correction amount.

[0008] Furthermore, the execution allocation unit adjusts the fuel injection quantity of the first-stage fuel injector according to the first-stage fuel injection correction quantity, adjusts the combustion gas injection quantity of the second-stage combustion gas injector according to the second-stage combustion gas injection correction quantity, and adjusts the shaping gas injection quantity of the third-stage flame shaping injector according to the third-stage shaping gas injection correction quantity.

[0009] Further, the process of generating heat flux density distribution and temperature distribution specifically includes the following steps: Step S1: The combustion space sub-model adopts a combustion reaction model that includes turbulence-chemical reaction interaction. Based on the fuel injection rate of the first-stage fuel injector, the combustion gas injection rate of the second-stage combustion gas injector, the shaping gas injection rate of the third-stage flame shaping injector, as well as the injection flow rate of each stage, combustion space temperature, flue gas oxygen content, flue gas carbon monoxide content, and furnace pressure, a flame heat release distribution is generated. The flame heat release distribution is used to characterize the heat release intensity and spatial distribution at the flame root, the middle and rear section of the flame, and the outer edge of the flame. Step S2: Based on the glass melt surface temperature, the corresponding glass melt viscosity parameters, glass melt thermal conductivity parameters, and glass melt specific heat parameters are matched to obtain the glass melt temperature-related physical properties. The position of the glass melt free surface is determined based on the glass melt level, and the position of the glass melt free surface is used as the free surface boundary. The glass melt sub-model is based on the glass melt surface temperature. For the liquid temperature-related properties and free surface boundaries, a temperature-viscosity laminar flow model is used to perform heat transfer weight allocation and temperature response iteration on adjacent computational units on the free surface of the molten glass, generating a temperature response distribution; Step S3: Map the flame heat release distribution to the combustion side of the thermal coupling interface, and map the temperature response distribution to the molten glass side of the thermal coupling interface; The thermal coupling interface corresponds to the molten glass free surface, and each interface computational unit on the thermal coupling interface corresponds to a free surface computational unit within the computational region of the molten glass free surface; Step S4: Perform bidirectional correspondence processing of heat flux density boundary information and temperature boundary information on the thermal coupling interface to obtain the heat flux density distribution and temperature distribution within the same computational region of the molten glass free surface; The heat flux density boundary information comes from the mapping result of the flame heat release distribution on the combustion side of the thermal coupling interface, and the temperature boundary information comes from the mapping result of the temperature response distribution on the molten glass side of the thermal coupling interface.

[0010] Further, step S1 specifically includes the following steps: Step S11: Construct a fuel injection boundary based on the fuel injection quantity of the first-stage fuel injector, construct an auxiliary combustion gas replenishment boundary based on the auxiliary combustion gas injection quantity of the second-stage combustion gas injector, and construct a shaping gas momentum boundary based on the shaping gas injection quantity of the third-stage flame shaping injector; Step S12: Determine the jet convergence region in the combustion space based on the fuel injection boundary, the auxiliary combustion gas replenishment boundary, and the shaping gas momentum boundary, and define the jet convergence region as the mixing region to be reacted; Step S13: Determine the local oxygen supply state, the local unburned component state, and the... Local reaction temperature state; Step S14: Based on the local oxygen supply state, local unburned component state, and local reaction temperature state, perform turbulent-chemical reaction interaction processing to determine the reaction lag state and local burnout state corresponding to the mixing area to be reacted; the reaction lag state indicates the area state where the oxygen-containing component and the unburned component are in effective contact conditions but have not completely undergone combustion reaction, and the local burnout state indicates the area state where the unburned component is below the preset burnout requirement; Step S15: Based on the reaction lag state and the local burnout state, determine the combustion reaction intensity distribution, and divide the flame root heat release area, the flame middle and rear section heat release area, and the flame outer edge heat release area according to the combustion reaction intensity distribution to generate the flame heat release distribution.

[0011] By adopting the above solution, the beneficial effects achieved by the present invention are as follows:

[0012] This invention establishes a thermal coupling interface corresponding to the free surface of the molten glass between the combustion space sub-model and the molten glass sub-model through a combustion-melting coupling prediction unit. This enables a unified expression of the heat flux density distribution on the combustion side and the temperature distribution on the molten glass side within the same calculation area of ​​the molten glass free surface. This improves the ability to identify the spatial correspondence between the flame heat input area and the molten glass temperature response area. It also solves the problem that existing furnace combustion control relies solely on furnace temperature, molten glass surface temperature, or flue gas composition, making it difficult to determine whether the heat input position matches the temperature response position. This enhances the accuracy and interpretability of judging the thermal state of local combustion stations in float glass furnaces.

[0013] This invention converts the positional difference between the heat flow concentration area and the temperature response area into longitudinal and lateral deviation values ​​through a co-positional deviation calculation unit. Combined with the glass melt surface temperature deviation, it forms the basis for injection correction, realizing the categorized and quantitative expression of flame heat release position deviation, flame lateral coverage deviation, and glass melt surface temperature deviation. This improves the correspondence between subsequent injection adjustment actions and specific deviation types, solves the problem in the prior art that fuel flow rate, combustion gas flow rate, or flame shape adjustment are easily coupled and the adjustment direction is unclear, and enhances the hierarchical and coordinated control capability between primary fuel injection, secondary combustion gas injection, and tertiary flame shaping injection.

[0014] This invention generates injection correction weights through the injection correction strategy model in the predictive control optimization unit, and performs constraint corrections within the feasible domain of injection correction defined by the furnace operating state parameters. This achieves coordinated allocation of injection correction amount between heat input adjustment requirements and furnace operating safety boundaries, improves the stability and executability of multi-level injection correction actions, and solves the problems of simultaneous large-scale adjustment of multiple injection levels, injection correction amount exceeding the valve execution range, or causing fluctuations in furnace pressure and flue gas burnout status. It enhances the combustion regulation stability, glass melt free surface temperature uniformity, and system closed-loop control reliability of float glass furnaces during continuous production. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall structure of a multi-stage combustion injection system for a float glass melting furnace as proposed in Embodiment 1 of the present invention;

[0016] Figure 2 This is a schematic diagram of the arrangement of the multi-stage injection unit in a combustion station as proposed in Embodiment 1 of the present invention;

[0017] Figure 3 This is a schematic diagram of the combustion-melting coupled prediction model structure proposed in Embodiment 1 of the present invention.

[0018] Explanation of reference numerals in the attached drawings: 100, float glass melting furnace; 110, combustion space; 120, molten glass; 121, free surface of molten glass; 130, melting section; 200, multi-stage injection unit; 210, primary fuel injector; 220, secondary combustion gas injector; 230, tertiary flame shaping injector; 231, first shaping injection branch; 232, second shaping injection branch; 310, molten glass surface temperature detector; 320, combustion space temperature detector; 330, flue gas composition detector; 340, furnace pressure detector; 350, molten glass level detector; 400, control unit; 410, combustion-melting coupling prediction unit; 420, co-position deviation calculation unit; 430, predictive control optimization unit; 440, execution allocation unit; 510, primary fuel flow regulating valve; 520, secondary combustion gas flow regulating valve; 530, tertiary shaping gas flow regulating valve. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In this specification, the length direction of the melting furnace refers to the direction from the feeding end to the refining end or the bottleneck direction; the width direction of the melting furnace refers to the direction that is perpendicular to the length direction of the melting furnace and parallel to the free surface 121 of the molten glass; the flame propagation direction refers to the extension direction of the main body of the flame in the combustion space 110 after the primary fuel injector 210 injects fuel; the free surface 121 of the molten glass is the surface area above the molten glass 120 adjacent to the combustion space 110, and the subsequent heat flux density distribution, temperature distribution, heat flux concentration area, temperature response area, longitudinal deviation value and lateral deviation value are all determined based on the free surface 121 of the molten glass.

[0021] Example 1, according to Figure 1 , Figure 2 and Figure 3 The present invention provides a multi-stage combustion injection system for a float glass melting furnace, which is applied to at least one combustion station in the melting section 130 of the float glass melting furnace 100. The combustion station can be set in the small furnace nozzle area on the side wall of the melting section 130, or it can be set in the auxiliary combustion injection position in the front area or hot spot area of ​​the melting section 130. The float glass melting furnace 100 includes a combustion space 110, a glass melt 120 located below the combustion space 110, and a glass melt free surface 121 located above the glass melt 120.

[0022] In this embodiment, the multi-stage combustion injection system includes a multi-stage injection unit 200, a detection unit (not shown in the figure), and a control unit 400; the multi-stage injection unit 200 is used to form a graded injection action within a combustion station; the detection unit is used to obtain furnace operating status parameters from field detectors and the production control system; the control unit 400 is used to generate injection correction actions based on the furnace operating status parameters and the current injection parameters of the multi-stage injection unit 200.

[0023] The multi-stage injection unit 200 includes a primary fuel injector 210, a secondary combustion gas injector 220, and a tertiary flame shaping injector 230. The primary fuel injector 210 is installed at the fuel injection position of the combustion station and is used to inject fuel into the combustion space 110 to form the flame root. The secondary combustion gas injector 220 is positioned behind the primary fuel injector 210 along the flame propagation direction and is used to supplement the combustion gas to the middle and rear sections of the flame. The tertiary flame shaping injector 230 is located at the outer edge of the flame and is used to inject shaping gas into the outer edge of the flame to adjust the lateral coverage of the flame on the free surface 121 of the molten glass.

[0024] In this embodiment, the fuel for the first-stage fuel injector 210 can be natural gas, coke oven gas, producer gas, or other commonly used fuels in glass melting furnaces; the fuel for the second-stage combustion-supporting injector 220 can be air, oxygen-enriched air, or oxygen; the shaping gas for the third-stage flame-shaping injector 230 can be preheated air, oxygen-enriched air, oxygen, low-oxygen flue gas, or a mixture of the above gases; the shaping gas is mainly used to change the momentum and lateral expansion range of the flame outer edge, and is not primarily intended to provide all the combustion oxygen.

[0025] The detection unit acquires furnace operating status parameters, including injection flow rates at each stage, combustion space temperature, molten glass surface temperature, flue gas oxygen content, flue gas carbon monoxide content, furnace pressure, and molten glass level. The injection flow rates at each stage reflect the current actual injection status of the primary fuel injector 210, the secondary combustion-supporting injector 220, and the tertiary flame-shaping injector 230, respectively. The combustion space temperature reflects the thermal state of the combustion space 110. The molten glass surface temperature reflects the heating result of the molten glass free surface 121. The flue gas oxygen content and flue gas carbon monoxide content reflect the burnout state. The furnace pressure constrains the injection adjustment actions. The molten glass level determines the position of the molten glass free surface 121.

[0026] The control unit 400 includes a combustion-melting coupling prediction unit, a co-position deviation calculation unit, a predictive control optimization unit, and an execution allocation unit; the detection unit and the multi-stage injection unit 200 are respectively connected to the control unit 400; the control unit 400 can be deployed in an edge computing controller and connected to the detection unit and the flow control valves at each stage via an industrial Ethernet.

[0027] like Figure 2 As shown, the combustion-melting coupled prediction unit includes a combustion space sub-model, a molten glass sub-model, and a thermal coupling interface. The combustion space sub-model receives furnace operating status parameters and current injection parameters to calculate the flame heat release distribution within the combustion space 110. The molten glass sub-model receives the molten glass surface temperature and molten glass level to calculate the temperature response distribution of the molten glass free surface 121. The thermal coupling interface 413 corresponds to the molten glass free surface 121 and is used to place the combustion-side heat input and the molten glass-side temperature response in the same molten glass free surface calculation region. The combustion space sub-model is responsible for calculating the heat generated by the flame; the molten glass sub-model is responsible for calculating the temperature change of the molten glass free surface. The two are connected by a thermal coupling interface. The thermal coupling interface can be understood as a virtual calculation interface corresponding to the molten glass free surface. It is not an actual installed mechanical component, but rather a computational boundary / data conversion layer in the control algorithm.

[0028] The combustion-melting coupled prediction unit 410 sets up a thermal coupling interface between the combustion space sub-model and the molten glass sub-model, and generates heat flux density distribution and temperature distribution within the same computational region of the molten glass free surface. The heat flux density distribution is used to characterize the location and intensity of heat input from the combustion space 110 to the molten glass free surface 121, and the temperature distribution is used to characterize the temperature response of the molten glass free surface 121 within the same computational region.

[0029] The colocation deviation calculation unit 420 receives the heat flux density distribution and temperature distribution, determines the heat flux concentration area based on the heat flux density distribution, determines the temperature response area based on the temperature distribution, and calculates the colocation deviation between the heat flux concentration area and the temperature response area. The colocation deviation includes the longitudinal deviation value along the length of the melting furnace and the lateral deviation value along the width of the melting furnace. The colocation deviation calculation unit 420 also determines the glass melt surface temperature deviation based on the difference between the temperature distribution and the preset target temperature distribution of the glass melt surface.

[0030] The predictive control optimization unit 430 includes an injection correction strategy model; the injection correction strategy model takes the longitudinal deviation value, the lateral deviation value and the glass melt surface temperature deviation as inputs to generate a first-level fuel injection correction weight, a second-level combustion gas injection correction weight and a third-level shaping gas injection correction weight; and further calculates the first-level fuel injection correction amount, the second-level combustion gas injection correction amount and the third-level shaping gas injection correction amount.

[0031] The execution allocation unit 440 receives the primary fuel injection correction amount, the secondary combustion gas injection correction amount, and the tertiary shaping gas injection correction amount, and converts them into flow setpoints or valve opening increments that can be executed by each injection branch. The execution allocation unit 440 adjusts the fuel injection amount of the primary fuel injector 210 according to the primary fuel injection correction amount, adjusts the combustion gas injection amount of the secondary combustion gas injector 220 according to the secondary combustion gas injection correction amount, and adjusts the shaping gas injection amount of the tertiary flame shaping injector 230 according to the tertiary shaping gas injection correction amount, so as to reduce the co-position deviation between the heat flow concentration area and the temperature response area.

[0032] like Figure 3 As shown, the specific arrangement of the multi-stage injection unit 200 and the corresponding deviation execution relationship are explained.

[0033] The injection outlet of the primary fuel injector 210 faces the combustion space 110, and the injection axis of the primary fuel injector 210 forms an angle with the free surface 121 of the molten glass, so that the fuel jet forms the root of the flame after entering the combustion space 110, instead of directly scouring the free surface 121 of the molten glass; the primary fuel injector 210 is connected to the fuel pipeline through the primary fuel flow regulating valve 510, which receives the fuel flow setting value issued by the execution distribution unit 440.

[0034] The injection outlet of the secondary combustion gas injector 220 is located downstream of the injection outlet of the primary fuel injector 210 along the flame propagation direction. The injection axis of the secondary combustion gas injector 220 intersects with the main flame axis formed by the primary fuel injector 210 at the middle or rear of the flame. The secondary combustion gas injector 220 is connected to the combustion gas pipeline through a secondary combustion gas flow regulating valve 520, which is used to execute the secondary combustion gas injection correction. Here, "downstream" does not refer to the downstream of the molten glass flow, but rather to the downstream of the flame propagation direction; that is, the direction in which the flame propagates forward after being ejected from the primary fuel injector 210. The secondary combustion gas injector 220 is a rear injector, not a flame root injector.

[0035] When the longitudinal deviation value indicates that the heat flow concentration area is ahead of the temperature response area along the length of the furnace, it means that the heat flow input position on the combustion side is too close to the upstream relative to the actual temperature response position of the molten glass. The predictive control optimization unit 430 increases the secondary combustion gas injection correction weight, and the execution allocation unit 440 increases the combustion gas replenishment amount in the middle and rear section of the flame according to the secondary combustion gas injection correction amount, so as to enhance the burnout and heat release in the middle and rear section of the flame.

[0036] When the longitudinal deviation value indicates that the heat flow concentration area lags behind the temperature response area along the length of the furnace, it means that the heat flow input position on the combustion side is too close to the downstream relative to the actual temperature response position of the molten glass. The predictive control optimization unit 430 reduces the secondary combustion gas injection correction weight or reduces the secondary combustion gas injection correction amount, thereby reducing the amount of combustion gas supplementation in the middle and rear sections of the flame and suppressing the heat release from continuing to shift towards the rear section of the flame.

[0037] The three-stage flame shaping injector 230 includes a first shaping injection branch 231 and a second shaping injection branch 232. The first shaping injection branch 231 and the second shaping injection branch 232 are respectively arranged on both sides of the first-stage fuel injector 210 and are used to inject shaping gas into the outer edges of both sides of the flame. The first shaping injection branch 231 and the second shaping injection branch 232 are both connected to the three-stage shaping gas flow regulating valve 530, and a first shaping gas flow regulating valve (not shown in the figure) and a second shaping gas flow regulating valve (not shown in the figure) are respectively provided. The three-stage shaping gas flow regulating valve 530 includes a first shaping gas flow regulating valve and a second shaping gas flow regulating valve. The shaping gas ejected from the two shaping injection branches acts on the outer edge of the flame to change the degree of lateral spread and lateral offset trend of the flame.

[0038] When the lateral deviation value indicates that the heat flow concentration area is biased to one side of the melting furnace width direction, the predictive control optimization unit 430 determines the differential shaping injection amount between the first shaping injection branch 231 and the second shaping injection branch 232 according to the lateral deviation value; the execution distribution unit 440 adjusts the first shaping gas flow regulating valve and the second shaping gas flow regulating valve according to the differential shaping injection amount, so that the momentum of the outer edge of the flame is redistributed in the melting furnace width direction, thereby correcting the lateral coverage range of the flame on the free surface 121 of the molten glass.

[0039] The process of the detection unit acquiring the furnace operating status parameters and the control unit 400 performing online calibration is explained.

[0040] The detection unit includes a glass melt surface temperature detector 310, a combustion space temperature detector 320, a flue gas composition detector 330, a melting furnace pressure detector 340, a glass melt level detector 350, and flow detectors corresponding to each level of injection branch (not shown in the figure); the above detectors can be connected to the production control system of the float glass melting furnace 100, or can be directly connected to the control unit 400.

[0041] The molten glass surface temperature detector 310 is used to output sampled values ​​of the molten glass surface temperature distributed along the length and width directions of the melting furnace; in this embodiment, the molten glass surface temperature detector 310 can be an infrared thermal imager. The control unit 400 maps the sampled values ​​of the molten glass surface temperature to corresponding calculation units within the calculation area of ​​the free surface of the molten glass, enabling them to be used for temperature distribution correction and determination of molten glass surface temperature deviation.

[0042] Combustion space temperature detector 320 is used to output the combustion space temperature; flue gas composition detector 330 is used to output the flue gas oxygen content and flue gas carbon monoxide content, wherein the flue gas oxygen content is used to determine the local oxygen supply status, and the flue gas carbon monoxide content is used to determine the local unburned component status; furnace pressure detector 340 is used to output furnace pressure, which is used to limit the influence of each stage of injection correction action on the furnace pressure balance; glass melt level detector 350 is used to output glass melt level, which is used to determine the position of the free surface of the glass melt.

[0043] Flow detectors are installed on the injection branches corresponding to the primary fuel injector 210, the secondary combustion-supporting injector 220, and the tertiary flame-shaping injector 230, respectively, to output the injection flow rate of each stage.

[0044] Example 2, based on Example 1, describes the process by which the combustion-melting coupling prediction unit 410 generates heat flux density distribution and temperature distribution. This process specifically includes the following steps:

[0045] Step S1: The combustion space sub-model 411 adopts a combustion reaction model that includes turbulence-chemical reaction interaction. Based on the fuel injection rate of the first-stage fuel injector 210, the combustion gas injection rate of the second-stage combustion gas injector 220, the shaping gas injection rate of the third-stage flame shaping injector 230, as well as the injection flow rate of each stage, the combustion space temperature, the flue gas oxygen content, the flue gas carbon monoxide content, and the furnace pressure, a flame heat release distribution is generated. The flame heat release distribution is used to characterize the heat release intensity and spatial distribution of the flame root, the middle and rear section of the flame, and the outer edge of the flame.

[0046] In step S1, the combustion space sub-model 411 can be implemented using a simplified jet mixing model, a computational fluid dynamics model, or a combustion reaction proxy model trained with historical operating data. The combustion space sub-model 411 first divides the combustion space 110 into multiple combustion space calculation units, and then configures the fuel injection boundary, combustion gas replenishment boundary, and shaping gas momentum boundary into the corresponding calculation units. Subsequently, jet convergence, local mixing, reaction hysteresis, and burnout state judgment are performed within the combustion space calculation units.

[0047] Step S2: Match the glass melt viscosity parameters, glass melt thermal conductivity parameters, and glass melt specific heat parameters according to the glass melt surface temperature to obtain the glass melt temperature-related properties; determine the glass melt free surface position according to the glass melt level, and use the glass melt free surface position as the free surface boundary of the glass melt sub-model 412; based on the glass melt temperature-related properties and free surface boundary, the glass melt sub-model 412 uses a temperature-viscosity laminar flow model to perform heat transfer weight allocation and temperature response iteration on adjacent computational units on the glass melt free surface 121 to generate the temperature response distribution of the glass melt free surface 121;

[0048] In step S2, the temperature-related physical properties of the molten glass can be determined by the glass composition, process formulation, and historical production data; the viscosity parameter of the molten glass is used to characterize the flow resistance of the molten glass at the corresponding temperature; the thermal conductivity parameter of the molten glass is used to characterize the heat transfer capability between adjacent computing units; and the specific heat parameter of the molten glass is used to characterize the temperature rise response capability of the free surface computing unit after being subjected to heat input.

[0049] In one specific embodiment, the glass melt sub-model 412 treats the glass melt 120 as a laminar fluid with temperature-dependent properties and uses it to solve for the mass conservation, momentum conservation, and energy conservation relationships of the glass melt 120; the correspondence between the glass melt viscosity and the glass melt temperature can be expressed as:

[0050] ;

[0051] in, The viscosity of the molten glass. The temperature of the molten glass. , and These are viscosity model coefficients determined based on the glass composition; for a specific float glass composition, , and The viscosity can be determined through a glass property database, measured viscosity data, or in-plant process data. The glass liquid sub-model 412 determines the flow resistance state corresponding to the free surface calculation unit based on the glass liquid viscosity, and generates the heat transfer weight between adjacent free surface calculation units by combining the glass liquid thermal conductivity parameters and glass liquid specific heat parameters, which is used for subsequent temperature response iteration.

[0052] Step S3: Map the flame heat release distribution to the combustion side of the thermal coupling interface 413, and map the temperature response distribution to the glass melt side of the thermal coupling interface 413; the thermal coupling interface 413 corresponds to the glass melt free surface 121, and each interface calculation unit on the thermal coupling interface 413 corresponds to a free surface calculation unit within the glass melt free surface calculation region.

[0053] Step S4: Perform bidirectional correspondence processing of heat flux density boundary information and temperature boundary information on the thermal coupling interface 413 to obtain the heat flux density distribution and temperature distribution within the same glass melt free surface calculation area; the heat flux density boundary information comes from the mapping result of the flame heat release distribution on the combustion side of the thermal coupling interface 413, and the temperature boundary information comes from the mapping result of the temperature response distribution on the glass melt side of the thermal coupling interface 413; since they use the same calculation area, the subsequent colocation deviation calculation unit 420 can directly compare the heat flux concentration area and the temperature response area in the same coordinate system.

[0054] Example 3, based on Example 2, describes the internal processing of generating the flame heat release distribution in the combustion space sub-model 411, specifically including the following steps:

[0055] Step S11: Construct a fuel injection boundary based on the fuel injection quantity of the first-stage fuel injector 210, a combustion gas replenishment boundary based on the combustion gas injection quantity of the second-stage combustion gas injector 220, and a shaping gas momentum boundary based on the shaping gas injection quantity of the third-stage flame shaping injector 230; the fuel injection boundary includes the fuel injection outlet position, fuel injection direction, fuel injection velocity, and fuel injection quantity; the combustion gas replenishment boundary includes the combustion gas injection outlet position, combustion gas injection direction, combustion gas injection velocity, and combustion gas injection quantity; the shaping gas momentum boundary includes the injection direction, injection velocity, and flow rate difference of the first shaping injection branch 231 and the second shaping injection branch 232;

[0056] Step S12: Based on the fuel injection boundary, combustion gas replenishment boundary, and shaping gas momentum boundary, determine the jet convergence area of ​​the fuel jet, combustion gas jet, and shaping gas jet in the combustion space 110, and define the jet convergence area as the reaction mixing area; the reaction mixing area is the common spatial basis for subsequent judgment of local oxygen supply state, local unburned component state, local reaction temperature state, reaction lag state, and local burnout state;

[0057] Step S13: Based on the combustion space temperature, flue gas oxygen content, and flue gas carbon monoxide content in the mixing zone to be reacted, determine the local oxygen supply state, local unburned component state, and local reaction temperature state corresponding to the mixing zone to be reacted; the local oxygen supply state is used to characterize whether the supply of oxygen-containing components in the mixing zone to be reacted meets the reaction requirements; the local unburned component state is used to characterize whether fuel components or incompletely burned components such as carbon monoxide are retained; the local reaction temperature state is used to characterize whether the region has reached the temperature conditions suitable for the combustion reaction to proceed.

[0058] Step S14: Based on the local oxygen supply status, the local unburned component status, and the local reaction temperature status, perform turbulence-chemical reaction interaction processing to determine the reaction lag status and local burnout status corresponding to the mixing area to be reacted; the reaction lag status indicates the area where the oxygen-containing component and the unburned component are in effective contact conditions but have not completely undergone combustion reaction, and the local burnout status indicates the area where the unburned component is below the preset burnout requirement.

[0059] Step S15: Determine the combustion reaction intensity distribution based on the reaction lag state and the partial burnout state, and divide the flame root heat release area, the middle and rear flame heat release area, and the outer edge flame heat release area according to the combustion reaction intensity distribution to generate the flame heat release distribution; the flame root heat release area mainly corresponds to the injection effect of the first-stage fuel injector 210, the middle and rear flame heat release area mainly corresponds to the oxygen supplementation effect of the second-stage combustion-supporting injector 220, and the outer edge flame heat release area mainly corresponds to the shaping gas momentum effect of the third-stage flame-shaping injector 230.

[0060] Example 4, based on Example 3, describes the process of the colocation deviation calculation unit 420 calculating the colocation deviation. This process specifically includes the following steps:

[0061] Step E1: Determine the preset heat flux threshold based on the heat flux density value in the heat flux density distribution, and determine the glass melt free surface calculation unit that reaches the preset heat flux threshold as the heat flux concentration area; In this embodiment, the preset heat flux threshold can be determined based on the highest preset proportion range in the heat flux density distribution, or it can be determined based on the heat flux density statistics in historical qualified production cycles.

[0062] Formula for determining the heat flux threshold:

[0063] ;

[0064] in, Indicates the first The heat flux threshold corresponding to each control cycle Indicates the first The average heat flux density of all free surface calculation cells in the free surface calculation region of the molten glass within one control cycle. Indicates the first Standard deviation of heat flux density of all free surface computational units within one control cycle The heat flux threshold adjustment coefficient has a value range of 0.2 to 2.0, preferably 0.5 to 1.2. Through this dynamic threshold method, the heat flux concentration area can be adaptively determined according to the overall level and dispersion of heat flux density in the current control cycle.

[0065] Step E2: Determine the preset temperature threshold based on the temperature values ​​in the temperature distribution, and determine the glass melt free surface calculation unit that reaches the preset temperature threshold as the temperature response region; In this embodiment, the preset temperature threshold can be determined based on the process target temperature corresponding to the glass type, or it can be determined based on the temperature values ​​within the highest preset proportion range in the current temperature distribution;

[0066] Formula for determining the temperature threshold:

[0067] ;

[0068] in, Indicates the first Temperature threshold corresponding to each control cycle; Indicates the first The average temperature of all free surface calculation units in the free surface calculation region of the molten glass within one control cycle; Indicates the first Temperature standard deviation of all free surface computational units within one control cycle; This represents the temperature threshold adjustment coefficient, used to adjust the screening range of the temperature response region. Through this dynamic threshold method, the temperature response region can be adaptively determined according to the overall level and dispersion of the free surface temperature of the molten glass within the current control cycle.

[0069] Step E3: Determine the heat flux weighting center of the heat flux concentration area based on the heat flux density value of each calculation unit in the heat flux concentration area, and determine the temperature weighting center of the temperature response area based on the temperature value of each calculation unit in the temperature response area; the heat flux weighting center is used to represent the main location of the heat input on the combustion side, and the temperature weighting center is used to represent the main location of the actual temperature response of the free surface 121 of the molten glass.

[0070] Step E4: Determine the longitudinal deviation value based on the coordinate difference between the heat flow weighted center and the temperature weighted center in the direction of the furnace length, and determine the lateral deviation value based on the coordinate difference between the heat flow weighted center and the temperature weighted center in the direction of the furnace width; the longitudinal deviation value is used to subsequently adjust the amount of combustion gas supplemented in the middle and rear section of the flame of the secondary combustion gas injector 220, and the lateral deviation value is used to subsequently adjust the shaping gas injection distribution of the tertiary flame shaping injector 230;

[0071] Step E5: Combine the longitudinal deviation value and the lateral deviation value into a colocation deviation between the heat flow concentration region and the temperature response region; the colocation deviation is different from the simple temperature deviation, which represents the positional correspondence between the heat input center on the combustion side and the temperature response center on the glass melt side; after the colocation deviation is generated, it is used by the predictive control optimization unit 430 to determine the injection correction weight.

[0072] Example 5, based on Example 4, includes a predictive control optimization unit comprising an injection correction strategy model. This model includes a sparse simplex weight generation layer and a polyhedral constraint safety correction layer. The sparse simplex weight generation layer uses longitudinal deviation, lateral deviation, and glass melt surface temperature deviation as inputs to generate initial weights for primary fuel injection, secondary combustion gas injection, and tertiary shaping gas injection, ensuring these weights satisfy non-negative normalization constraints. The polyhedral constraint safety correction layer constructs a feasible injection correction domain based on furnace operating parameters and performs constraint corrections on the initial weights within this domain to obtain the corrected weights for primary fuel injection, secondary combustion gas injection, and tertiary shaping gas injection.

[0073] The process by which the sparse simplex weight generation layer generates the initial weights for the first-stage fuel injection, the initial weights for the second-stage combustion gas injection, and the initial weights for the third-stage shaping gas injection specifically includes the following steps:

[0074] Step A1: Construct the injection correction demand vector; Perform amplitude normalization on the longitudinal deviation value, lateral deviation value, and glass melt surface temperature deviation respectively to obtain the normalized values ​​of longitudinal deviation, lateral deviation, and temperature deviation. Construct the injection correction demand vector according to the mapping relationship of temperature deviation correction corresponding to primary fuel injection, longitudinal deviation correction corresponding to secondary combustion gas injection, and lateral deviation correction corresponding to tertiary shaping gas injection.

[0075] Step A2: Construct a uniform prior candidate weight pool; generate multiple disturbance demand vectors based on the injection correction demand vector, and input the multiple disturbance demand vectors into multiple weight candidate generators respectively; in the initial state, assign the same prior selection probability to each weight candidate generator, so that different weight candidate generators have a uniform prior distribution when entering the current control cycle; each weight candidate generator outputs a set of candidate injection weights, which include primary fuel candidate weights, secondary combustion gas candidate weights, and tertiary shaping gas candidate weights;

[0076] Step A3: Perform simplex constraint transformation; perform nonnegation and normalization processing on each group of candidate injection weights so that each group of candidate injection weights satisfies that the first-stage fuel candidate weight, the second-stage combustion gas candidate weight, and the third-stage shaping gas candidate weight are all not less than zero, and the sum of the three is equal to one; thereby restricting each group of candidate injection weights to the three-dimensional injection weight simplex composed of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection.

[0077] Step A4: Perform weight update from uniform prior to sparse posterior; calculate the posterior score of each weight candidate generator based on the matching degree between each group of candidate injection weights and the current injection correction demand vector, as well as the reduction in colocation deviation and the reduction in glass melt surface temperature deviation caused by the corresponding candidate injection weights in the historical control cycle; reduce the posterior selection probability of weight candidate generators with posterior scores lower than the preset sparse screening requirements, or set their posterior selection probability to zero; retain the posterior selection probability of weight candidate generators with posterior scores that meet the preset sparse screening requirements, thereby transforming the uniform prior distribution into a sparse posterior distribution;

[0078] Step A5: Perform sparse Bagging integration output; based on the sparse posterior distribution, perform posterior weighted integration on the retained multiple sets of candidate injection weights to obtain the initial weights of primary fuel injection, secondary combustion gas injection, and tertiary shaping gas injection; wherein the initial weights of primary fuel injection, secondary combustion gas injection, and tertiary shaping gas injection satisfy the non-negative normalization constraint, and in the current control cycle, the injection level corresponding to the dominant deviation obtains a higher initial weight.

[0079] The process by which the polyhedral constraint safety correction layer constrains and corrects the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection specifically includes the following steps:

[0080] Step C1: Construct the injection correction state vector; extract the flue gas oxygen content, flue gas carbon monoxide content, furnace pressure, injection flow rate at each stage, and valve position feedback of the flow control valve at each stage from the furnace operating state parameters, and combine them with the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection to obtain the injection correction state vector.

[0081] Step C2: Construct a polyhedral injection correction feasible region; Based on the upper and lower limits of the first-stage fuel injection quantity, the upper and lower limits of the second-stage combustion gas injection quantity, the upper and lower limits of the third-stage shaping gas injection quantity, the allowable fluctuation range of furnace pressure, the allowable range of flue gas oxygen content, the allowable range of flue gas carbon monoxide content, and the upper limit of valve single-cycle change amplitude, generate multiple linear constraint boundaries, and form an injection correction feasible region by enclosing multiple linear constraint boundaries;

[0082] Step C3: Perform constraint embedding processing; embed the constraint boundary of the injection correction feasible region into the weight update process of the polyhedral constraint safety correction layer, so that the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection simultaneously satisfy the non-negative normalization constraint, injection flow constraint, furnace pressure constraint, and flue gas burnout constraint during the update process; wherein, the flue gas burnout constraint is determined based on the flue gas oxygen content and flue gas carbon monoxide content;

[0083] Step C4: Perform a non-projection constraint update; within the feasible region of injection correction, based on the correction requirements corresponding to the current longitudinal deviation value, lateral deviation value, and glass melt surface temperature deviation, iteratively update the initial weights of the first-stage fuel injection, the initial weights of the second-stage combustion gas injection, and the initial weights of the third-stage shaping gas injection along the direction inside the feasible region to obtain candidate correction weights; the candidate correction weights remain within the feasible region of injection correction during the iterative update process, without needing to perform projection correction from the weights outside the feasible region to the weights inside the feasible region after the iteration is completed.

[0084] Step C5: Output safety correction weights; perform safety boundary verification on candidate correction weights, and determine the candidate correction weights as first-level fuel injection correction weights, second-level combustion gas injection correction weights, and third-level shaping gas injection correction weights if the candidate correction weights meet the feasible region of the injection correction; when a candidate correction weight approaches any constraint boundary, reduce the weight update magnitude of the corresponding injection level in the next control cycle.

[0085] Example 6, based on Example 5, uses a combustion station in the front zone of the hot spot of the melting section 130 of the float glass melting furnace 100 as an example to illustrate the complete operation process of the multi-stage combustion injection system; the values ​​in this example are used to illustrate the connection between the various technical links and do not represent a limitation on the scope of protection of the present invention.

[0086] In this embodiment, the control cycle of the control unit 400 is 30s; the combustion station is located on the side wall of the melting section 130 of the float glass melting furnace 100, and the calculation area corresponding to the free surface 121 of the molten glass is 4.0m along the length direction of the melting furnace and 2.4m along the width direction of the melting furnace, and is divided into 8×6 free surface calculation units; the x-coordinate of the length direction of the melting furnace is positively oriented with the direction from the feeding end to the refining end, and the y-coordinate of the width direction of the melting furnace is positively oriented with the direction from the side wall where the combustion station is located to the center line of the melting furnace.

[0087] At the start of the kth control cycle, the detection unit acquires the current furnace operating status parameters; among them, the fuel for the first-stage fuel injector 210 is natural gas, and the current fuel injection rate is 280 Nm³ / h; the combustion gas for the second-stage combustion-supporting injector 220 is oxygen-enriched air, and the current combustion-supporting gas injection rate is 780 Nm³ / h; the shaping gas for the third-stage flame-shaping injector 230 is preheated air, with the current shaping gas injection rate of the first shaping injection branch 231 being 130 Nm³ / h, and the current shaping gas injection rate of the second shaping injection branch 232 being 130 Nm³ / h. The combustion space temperature detector 320 detected a combustion space temperature of 1585℃ above the combustion station; the glass melt surface temperature detector 310 detected a temperature sampling value of 1536℃ to 1561℃ on the free surface 121 of the glass melt; the flue gas composition detector 330 detected an oxygen content of 2.1% and a carbon monoxide content of 220ppm in the flue gas; the furnace pressure detector 340 detected a furnace pressure of +1.5Pa; and the glass melt level detector 350 detected a glass melt level of +3mm relative to the reference level.

[0088] The combustion-melting coupling prediction unit 410 inputs the above-mentioned furnace operating status parameters and current injection parameters into the combustion space sub-model 411 and the glass melt sub-model 412. The combustion space sub-model 411 constructs the fuel injection boundary, the combustion gas replenishment boundary, and the shaping gas momentum boundary based on the fuel injection amount of the first-stage fuel injector 210, the combustion gas injection amount of the second-stage combustion gas injector 220, and the shaping gas injection amount of the first shaping injection branch 231 and the second shaping injection branch 232, and determines the jet convergence area of ​​the fuel jet, the combustion gas jet, and the shaping gas jet in the combustion space 110. The combustion space sub-model 411 determines the local oxygen supply state, the local unburned component state, and the local reaction temperature state based on the combustion space temperature, flue gas oxygen content, and flue gas carbon monoxide content in the jet convergence area, and generates the flame heat release distribution after turbulence-chemical reaction interaction processing.

[0089] During this control cycle, the flame heat release distribution output by the combustion space sub-model 411 shows that the relative heat release intensity corresponding to the heat release region at the flame root is 0.82, the relative heat release intensity corresponding to the heat release region in the middle and rear of the flame is 0.66, and the relative heat release intensity corresponding to the heat release region at the outer edge of the flame is 0.74 on the positive side of the furnace width and 0.58 on the negative side of the furnace width. This result indicates that the heat release in the middle and rear of the flame is relatively insufficient, and the heat release at the outer edge of the flame is offset along the positive direction of the furnace width.

[0090] The glass melt sub-model 412 determines the position of the glass melt free surface 121 based on the sampled glass melt surface temperature and glass melt level, and obtains the glass melt viscosity parameters, glass melt thermal conductivity parameters, and glass melt specific heat parameters based on the glass composition. The glass melt sub-model 412 divides the glass melt free surface 121 into multiple free surface calculation units, and performs heat transfer weight allocation and temperature response iteration based on the temperature difference between adjacent calculation units, the flow resistance state corresponding to viscosity, thermal conductivity parameters, and specific heat parameters to obtain the temperature response distribution of the glass melt free surface 121.

[0091] The combustion-melting coupling prediction unit 410 maps the flame heat release distribution to the combustion side of the thermal coupling interface 413 and the temperature response distribution to the glass melt side of the thermal coupling interface 413, obtaining the heat flux density distribution and temperature distribution within the same glass melt free surface calculation region. In this control cycle, the average heat flux density of each free surface calculation unit on the glass melt free surface 121 is 112.0 kW / m², with a standard deviation of 18.0 kW / m²; the average temperature of each free surface calculation unit is 1548.5℃, with a standard deviation of 8.0℃.

[0092] The colocation deviation calculation unit 420 determines the heat flux concentration area based on the heat flux density distribution. In this embodiment, the heat flux threshold adjustment coefficient is set to 0.8, so the heat flux threshold is 126.4 kW / m².

[0093] The co-location deviation calculation unit 420 defines the free surface calculation unit with a heat flux density of not less than 126.4 kW / m² as the heat flux concentration region. Simultaneously, the temperature threshold adjustment coefficient is set to 0.8, resulting in a temperature threshold of 1554.9℃.

[0094] The colocation deviation calculation unit 420 defines the free surface calculation unit with a temperature value of not less than 1554.9℃ as the temperature response region.

[0095] Subsequently, the colocation deviation calculation unit 420 calculates the heat flux weighted center based on the heat flux density values ​​of each free surface calculation unit within the heat flux concentration region, and calculates the temperature weighted center based on the temperature values ​​of each free surface calculation unit within the temperature response region. In this control cycle, the heat flux weighted center is (5.20m, 0.38m), and the temperature weighted center is (5.55m, -0.12m). Therefore, the longitudinal deviation value is -0.35m; the lateral deviation value is 0.50m.

[0096] Among them, a longitudinal deviation value <0 indicates that the heat flow concentration area is offset upstream relative to the temperature response area along the length of the melting furnace, indicating that the current flame heat release position is forward relative to the glass melt temperature response position; a lateral deviation value >0 indicates that the heat flow concentration area is offset towards the positive direction of the melting furnace width relative to the temperature response area, indicating that the flame lateral coverage range is biased to one side; the co-position deviation calculation unit 420 also determines the glass melt surface temperature deviation to be 6.5℃ based on the difference between the temperature distribution and the preset target temperature distribution, indicating that the current glass melt free surface temperature is generally lower than the target temperature level.

[0097] The predictive control optimization unit 430 inputs the longitudinal deviation value, the lateral deviation value, and the glass melt surface temperature deviation into the injection correction strategy model. The sparse simplex weight generation layer normalizes the amplitude of the above deviations with the upper limit of longitudinal deviation (0.80m), the upper limit of lateral deviation (0.60m), and the upper limit of temperature deviation (20℃), respectively, resulting in a normalized value of 0.438 for longitudinal deviation, 0.833 for lateral deviation, and 0.325 for temperature deviation. Based on this, an injection correction demand vector [0.325, 0.438, 0.833] is constructed, where the first term corresponds to the first-level fuel injection correction demand, the second term corresponds to the second-level combustion gas injection correction demand, and the third term corresponds to the third-level shaping gas injection correction demand.

[0098] In this embodiment, the sparse simplex weight generation layer is configured with 5 weight candidate generators; the 5 weight candidate generators output candidate injection weights W1(k)=[0.22, 0.31, 0.47], W2(k)=[0.18, 0.29, 0.53], W3(k)=[0.25, 0.20, 0.55], W4(k)=[0.33, 0.38, 0.29], and W5(k)=[0.28, 0.26, 0.46] respectively in the current control cycle; the above candidate injection weights are all subjected to simplex constraint transformation so that each weight is not less than zero and the sum of the three is equal to one.

[0099] The sparse simplex weight generation layer obtains the posterior scores of five weight candidate generators based on the matching degree between the candidate injection weights and the injection correction demand vector, as well as the reduction in colocation deviation and the reduction in glass melt surface temperature deviation caused by the corresponding candidate injection weights in the historical control cycle. After sparse posterior update, the posterior selection probability of the fourth weight candidate generator is set to zero because it is lower than the preset sparse screening requirement. The posterior selection probabilities of the remaining candidate generators are 0.28, 0.35, 0.22, and 0.15, respectively. The sparse simplex weight generation layer performs posterior weighted integration on the retained candidate injection weights according to the sparse posterior distribution to obtain the initial weights of the first-stage fuel injection (0.222), the second-stage combustion gas injection (0.271), and the third-stage shaping gas injection (0.507).

[0100] The polyhedral constraint safety correction layer constructs a feasible injection correction domain based on the furnace operating status parameters. In this embodiment, the allowable fuel injection quantity variation range of the primary fuel injector 210 within a single control cycle is -8 Nm³ / h to +8 Nm³ / h; the allowable combustion gas injection quantity variation range of the secondary combustion gas injector 220 within a single control cycle is -50 Nm³ / h to +50 Nm³ / h; the allowable single-branch shaping gas injection quantity variation range of the tertiary flame shaping injector 230 within a single control cycle is -20 Nm³ / h to +20 Nm³ / h; the allowable furnace pressure range is -3 Pa to +5 Pa; the allowable flue gas oxygen content range is 1.5% to 4.0%; the upper limit of flue gas carbon monoxide content is 500 ppm; and the single-cycle valve position change amplitude of each flow regulating valve does not exceed 4%.

[0101] The polyhedral constraint safety correction layer embeds the aforementioned constraint boundaries into the weight update process and performs unprojected constraint updates on the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection within the feasible region of injection correction. Since the current furnace pressure is +1.5 Pa, which is lower than the upper limit of +5 Pa, and the flue gas oxygen content is 2.1% and the flue gas carbon monoxide content is 220 ppm, both within the allowable range, a moderate increase in the first-stage fuel injection quantity and the second-stage combustion gas injection quantity is permitted. At the same time, since the lateral deviation value is >0, the polyhedral constraint safety correction layer increases the third-stage shaping gas injection correction weight and allocates the third-stage shaping gas correction mainly to the differential injection quantity between the first shaping injection branch 231 and the second shaping injection branch 232. After constraint correction, the first-stage fuel injection correction weight is 0.200, the second-stage combustion gas injection correction weight is 0.280, and the third-stage shaping gas injection correction weight is 0.520.

[0102] The predictive control optimization unit 430 calculates the injection correction amount based on the above correction weights. In this control cycle, the first-level fuel injection correction amount is +4 Nm³ / h, the second-level combustion gas injection correction amount is +28 Nm³ / h, and the third-level shaping gas injection correction amount is manifested as a decrease of 12 Nm³ / h in the first shaping injection branch 231 and an increase of 12 Nm³ / h in the second shaping injection branch 232. The above-mentioned shaping gas differential injection amount is used to redistribute the momentum of the outer edge of the flame in the negative direction of the furnace width, thereby weakening the tendency of the heat flow concentration area to shift in the positive direction of the furnace width.

[0103] The execution allocation unit 440 receives the injection correction amount output by the predictive control optimization unit 430 and converts it into the set value of the corresponding flow control valve. The execution allocation unit 440 outputs a fuel flow set value of 284 Nm³ / h to the first-stage fuel flow control valve 510, an auxiliary combustion flow set value of 808 Nm³ / h to the second-stage combustion gas flow control valve 520, a shaping gas flow set value of 118 Nm³ / h to the first shaping gas flow control valve in the third-stage shaping gas flow control valve 530, and a shaping gas flow set value of 142 Nm³ / h to the second shaping gas flow control valve in the third-stage shaping gas flow control valve 530. The execution allocation unit 440 simultaneously reads the valve position feedback and actual flow feedback of each flow control valve and uses the valve position feedback and actual flow feedback as input data for the next control cycle.

[0104] After executing the above control for 10 consecutive control cycles, the detection unit re-acquires the furnace operating status parameters. At this time, the actual fuel injection rate of the primary fuel injector 210 is stable at 284 Nm³ / h, the actual combustion gas injection rate of the secondary combustion gas injector 220 is stable at 808 Nm³ / h, the actual shaping gas injection rate of the first shaping injection branch 231 is 118 Nm³ / h, and the actual shaping gas injection rate of the second shaping injection branch 232 is 142 Nm³ / h. The combustion space temperature is 1592℃, the flue gas oxygen content is 2.3%, the flue gas carbon monoxide content is 160 ppm, and the furnace pressure is +2.1 Pa, which still meets the constraint requirements.

[0105] The heat flux density distribution and temperature distribution generated by the combustion-melting coupling prediction unit 410 in the new control cycle show that the heat flux weighting center is adjusted to (5.43m, 0.12m), the temperature weighting center is adjusted to (5.55m, -0.03m), the longitudinal deviation value decreases from -0.35m to -0.12m, and the lateral deviation value decreases from 0.50m to 0.15m; the average temperature of the free surface 121 of the molten glass increases from 1548.5℃ to 1552.1℃, and the temperature deviation of the molten glass surface decreases from 6.5℃ to 2.9℃. Simultaneously, the overlap between the heat flux concentration area and the temperature response area increases from 0.38 to 0.62. These results indicate that through the graded correction of the primary fuel injection quantity, the secondary combustion-supporting gas injection quantity, and the tertiary flame shaping injection quantity, the spatial matching degree between the flame heat input area and the molten glass temperature response area is improved, and the flue gas burnout state and furnace pressure of the combustion space 110 remain within the allowable range.

[0106] The complete operation process of this embodiment shows that the combustion-melting coupling prediction unit 410 can generate heat flux density distribution and temperature distribution within the same glass melt free surface calculation area; the colocation deviation calculation unit 420 can convert the positional difference between the heat flux concentration area and the temperature response area into longitudinal deviation value and lateral deviation value; the predictive control optimization unit 430 can use the injection correction strategy model to generate injection correction weights that satisfy non-negative normalization constraints and polyhedral constraints; and the execution allocation unit 440 can perform graded execution of primary fuel injection, secondary combustion gas injection and tertiary flame shaping injection according to the injection correction amount, thereby forming a closed-loop adjustment process from detection, prediction, deviation calculation, strategy optimization to execution feedback.

[0107] The present invention and its embodiments have been described above. This description is not restrictive. The accompanying drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the present invention, such design should fall within the protection scope of the present invention.

Claims

1. A multi-stage combustion injection system for a float glass melting furnace, applied to a combustion station in the melting section of the float glass melting furnace, the float glass melting furnace comprising a combustion space and a free surface of molten glass, characterized in that, The system includes: The multi-stage injection unit includes a primary fuel injector, a secondary combustion-supporting injector, and a tertiary flame-shaping injector. The detection unit acquires the operating status parameters of the melting furnace; The control unit includes a combustion-melting coupling prediction unit, a co-position deviation calculation unit, a predictive control optimization unit, and an execution allocation unit; The combustion-melting coupling prediction unit includes a combustion space sub-model, a glass melt sub-model, and a thermal coupling interface. It takes the furnace operating status parameters and the current injection parameters of the first-stage fuel injector, the second-stage combustion-supporting injector, and the third-stage flame-shaping injector as inputs. A thermal coupling interface corresponding to the free surface of the glass melt is set between the combustion space sub-model and the glass melt model to generate heat flux density distribution and temperature distribution. The colocation deviation calculation unit calculates the colocation deviation based on the heat flux density distribution and temperature distribution; and determines the surface temperature deviation of the molten glass based on the difference between the temperature distribution and the preset target temperature distribution. The predictive control optimization unit includes a jetting correction strategy model; the jetting correction strategy model takes the colocation deviation and the surface temperature deviation of the molten glass as inputs, generates jetting correction weights, and further calculates the jetting correction amount; The distribution unit adjusts the fuel injection quantity of the first-stage fuel injector, the combustion gas injection quantity of the second-stage combustion-supporting injector, and the shaping gas injection quantity of the third-stage flame-shaping injector according to the injection correction amount.

2. The multi-stage combustion injection system for a float glass melting furnace according to claim 1, characterized in that: The furnace operating parameters include the flow rate of each injection stage, the temperature of the combustion space, the surface temperature of the molten glass, the oxygen content of the flue gas, the carbon monoxide content of the flue gas, the furnace pressure, and the molten glass level.

3. A multi-stage combustion injection system for a float glass melting furnace according to claim 1, characterized in that: The colocation deviation calculation unit determines the heat flux concentration area based on the heat flux density distribution, determines the temperature response area based on the temperature distribution, and calculates the colocation deviation between the heat flux concentration area and the temperature response area. The colocation deviation includes the longitudinal deviation value in the length direction of the melting furnace and the lateral deviation value in the width direction of the melting furnace. Based on the difference between the temperature distribution and the preset target temperature distribution, the surface temperature deviation of the molten glass is determined.

4. A multi-stage combustion injection system for a float glass melting furnace according to claim 3, characterized in that: The predictive control optimization unit includes a jet correction strategy model; the jet correction strategy model includes a sparse simplex weight generation layer and a polyhedral constraint safety correction layer. The sparse simplex weight generation layer takes the longitudinal deviation value, the transverse deviation value and the glass melt surface temperature deviation as inputs to generate the initial weights of the first-stage fuel injection, the initial weights of the second-stage combustion gas injection and the initial weights of the third-stage shaping gas injection, and makes the initial weights of the first-stage fuel injection, the initial weights of the second-stage combustion gas injection and the initial weights of the third-stage shaping gas injection satisfy the non-negative normalization constraint. The polyhedral constraint safety correction layer constructs an injection correction feasible region based on the furnace operating status parameters, and performs constraint correction on the initial weights of the first-stage fuel injection, the second-stage combustion gas injection, and the third-stage shaping gas injection within the injection correction feasible region to obtain the first-stage fuel injection correction weights, the second-stage combustion gas injection correction weights, and the third-stage shaping gas injection correction weights. Furthermore, the first-stage fuel injection correction, the second-stage combustion gas injection correction, and the third-stage shaping gas injection correction are calculated.

5. A multi-stage combustion injection system for a float glass melting furnace according to claim 4, characterized in that: The execution allocation unit adjusts the fuel injection quantity of the first-stage fuel injector according to the first-stage fuel injection correction quantity, adjusts the combustion gas injection quantity of the second-stage combustion gas injector according to the second-stage combustion gas injection correction quantity, and adjusts the shaping gas injection quantity of the third-stage flame shaping injector according to the third-stage shaping gas injection correction quantity.

6. A multi-stage combustion injection system for a float glass melting furnace according to claim 2, characterized in that: The process of generating heat flux density distribution and temperature distribution specifically includes the following steps: Step S1: The combustion space sub-model adopts a combustion reaction model that includes turbulence-chemical reaction interaction. Based on the fuel injection rate of the first-stage fuel injector, the combustion gas injection rate of the second-stage combustion gas injector, the shaping gas injection rate of the third-stage flame shaping injector, as well as the injection flow rate of each stage, the combustion space temperature, the flue gas oxygen content, the flue gas carbon monoxide content, and the furnace pressure, the flame heat release distribution is generated. Step S2: Based on the surface temperature of the molten glass, match the corresponding viscosity parameters, thermal conductivity parameters, and specific heat parameters of the molten glass to obtain the temperature-related physical properties of the molten glass; determine the position of the free surface of the molten glass based on the liquid level of the molten glass, and use the position of the free surface of the molten glass as the boundary of the free surface; the molten glass sub-model, based on the temperature-related physical properties of the molten glass and the boundary of the free surface, uses a temperature-viscosity laminar flow model to perform heat transfer weight allocation and temperature response iteration on adjacent computational units on the free surface of the molten glass to generate a temperature response distribution; Step S3: Map the flame heat release distribution to the combustion side of the thermal coupling interface, and map the temperature response distribution to the molten glass side of the thermal coupling interface; Step S4: Perform bidirectional correspondence processing of heat flux density boundary information and temperature boundary information on the thermal coupling interface to obtain the heat flux density distribution and temperature distribution within the same glass melt free surface calculation region.

7. A multi-stage combustion injection system for a float glass melting furnace according to claim 6, characterized in that: Step S1 specifically includes the following steps: Step S11: Construct the fuel injection boundary based on the fuel injection quantity of the first-stage fuel injector, construct the combustion gas replenishment boundary based on the combustion gas injection quantity of the second-stage combustion gas injector, and construct the shaping gas momentum boundary based on the shaping gas injection quantity of the third-stage flame shaping injector. Step S12: Based on the fuel injection boundary, combustion gas replenishment boundary, and shaping gas momentum boundary, determine the jet convergence region in the combustion space, and define the jet convergence region as the mixing region to be reacted; Step S13: Determine the local oxygen supply status, local unburned component status, and local reaction temperature status corresponding to the mixing area to be reacted; Step S14: Based on the local oxygen supply status, the local unburned component status, and the local reaction temperature status, perform turbulence-chemical reaction interaction processing to determine the reaction lag status and local burnout status corresponding to the mixing region to be reacted. Step S15: Determine the combustion reaction intensity distribution based on the reaction lag state and the local burnout state, and divide the heat release area at the flame root, the heat release area in the middle and rear section of the flame, and the heat release area at the outer edge of the flame according to the combustion reaction intensity distribution to generate the flame heat release distribution.

Images