Method, device and equipment for optimizing flow field of denitration system of ultra-supercritical unit and storage medium

Abstract

The invention belongs to the technical field of flue gas denitration, and discloses an ultra-supercritical unit denitration system flow field optimization method, device and equipment and a storage medium, and the method comprises the steps: determining the position of a drift diameter separation guide plate in a trapezoidal variable-diameter flue through employing a denitration flue gas dynamic flow field model established based on a flue structure of a target denitration system, and carrying out the installation; the flue gas flow field distribution in the flue is optimized; the concentration of NOx in the flue gas is measured in different areas in real time in the optimized flue based on a multi-point mixing method; based on the output of an inlet NOx concentration real-time prediction model constructed by utilizing the NOx concentration data measured in real time in the partition, constructing an ammonia injection total amount prediction feedforward control algorithm; by means of the algorithm, the three-level valves are arranged in series, digital-intelligent cooperative control and adjustment of the ammonia spraying flow are achieved, the distribution uniformity of a flue gas flow field in a flue can be effectively improved, the NOx concentration measurement precision is improved, real-time accurate adjustment of the ammonia spraying amount is achieved, and therefore the denitration efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of flue gas denitrification technology, and in particular to a method, apparatus, equipment and storage medium for optimizing the flow field of a denitrification system for an ultra-supercritical unit. Background Technology

[0002] With continuously rising environmental standards, the SCR denitrification systems of ultra-supercritical coal-fired power plants face severe challenges during operation. Under deep peak-shaving or low-load conditions, the uneven flow field distribution caused by design flaws in the flue gas duct structure is particularly prominent. This is especially true in trapezoidal variable-diameter flue gas ducts, where eddies and low-velocity zones easily form, leading to an imbalance in flue gas velocity distribution. This uneven flow field directly affects the mixing effect of flue gas and reducing agent, reducing the reaction rate on the catalyst surface. Insufficient denitrification occurs in some areas while ammonia is excessive in others, causing localized ammonia escape and the formation of ammonium bisulfate, which in turn exacerbates the risk of air preheater blockage and equipment corrosion. Current technologies rely heavily on empirically installed guide vanes for flow field optimization, lacking precise positioning and parameter design based on dynamic flow field models, making it difficult to adapt to complex and variable operating conditions. Furthermore, NOx concentration measurement generally uses single-point sampling, failing to capture the regional concentration variation characteristics of the flue gas duct cross-section, resulting in insufficient data for ammonia injection control. Ammonia injection systems often employ traditional feedback control strategies, whose response speed lags behind load fluctuations, making it difficult to achieve real-time and precise adjustment of the ammonia injection rate. This leads to fluctuations in denitrification efficiency and increased ammonia slip during load changes. Although some studies have attempted to improve the flow field through structural modifications, they have failed to organically combine flow field optimization, zoned concentration measurement, and intelligent predictive control. The lack of quantitative analysis and coordinated regulation of flue gas dynamics hinders the stable and efficient operation of denitrification systems under complex conditions.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this invention is to provide a method, apparatus, equipment, and storage medium for optimizing the flow field of an ultra-supercritical unit denitrification system. This invention aims to solve the technical problem that the failure to organically combine flow field optimization, zoned concentration measurement, and intelligent predictive control, as well as the lack of quantitative analysis and coordinated regulation capabilities of flue gas dynamic characteristics, restricts the stable and efficient operation of the denitrification system under complex operating conditions, thereby reducing denitrification efficiency.

[0005] To achieve the above objectives, the present invention provides a method for optimizing the flow field of a denitrification system in an ultra-supercritical unit, the method comprising the following steps: A dynamic flow field model of denitrification flue gas was established using simulation software based on the flue structure of the target denitrification system. Based on the dynamic flow field model, the location for installing the flow divider baffle in the trapezoidal variable diameter flue was determined and the baffle baffle was installed to optimize the flue gas flow field distribution in the flue. Within the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Using real-time NOx concentration data measured in different zones, a real-time prediction model for ingress NOx concentration is constructed by combining the Transformer model with ensemble learning. A feedforward control algorithm for predicting total ammonia injection volume is constructed based on the output of the real-time prediction model of the inlet NOx concentration. Based on the feedforward control algorithm for total ammonia injection prediction, a three-stage valve system consisting of a total ammonia injection automatic control valve, a zoned ammonia injection automatic regulating valve, and a branch pipe regulating manual valve is connected in series to achieve intelligent and coordinated control and regulation of ammonia injection flow.

[0006] In one embodiment, determining and installing the diameter-separating guide plate within the trapezoidal variable-diameter flue based on the dynamic flow field model to optimize the flue gas flow field distribution within the flue includes: Based on the dynamic flow field model, the initial flow field in the trapezoidal variable diameter flue is simulated and analyzed. The simulation analysis is used to identify key locations where the flow field distribution is uneven, vortices exist, or low-speed zones are present. Based on the simulation analysis results, various arrangement schemes for the flow dividers are determined. The arrangement schemes include the number of flow dividers, installation positions, installation angles, and geometric parameters. In the dynamic flow field model, each arrangement scheme of the flow divider is loaded and simulation calculations are performed. Comparative analysis of the simulation results of each scheme was conducted, and the evaluation indicators included the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change in system drag. Based on the preset optimization target, a target arrangement scheme is selected from various arrangement schemes of the flow divider baffles. The target arrangement scheme is the flow divider baffle arrangement scheme with the best overall performance. Based on the design parameters of the target layout scheme, a flow divider plate is installed at the actual physical location of the trapezoidal variable diameter flue to optimize the flue gas flow field distribution within the flue.

[0007] In one embodiment, the step of performing real-time zoned measurement of NOx concentration in flue gas using a multi-point mixing method includes: Based on the optimized flow field characteristics of the flue, the cross-section of the flue is divided into multiple logical measurement regions; Multiple NOx concentration sampling probes are deployed within each logical measurement area to form a distributed sampling network; The distributed sampling network mixes the flue gas from each probe within each logical measurement area proportionally to obtain a mixed flue gas sample, which is used to represent the average concentration of the corresponding logical measurement area. Using a continuous emission monitoring system or rapid analyzer, the mixed flue gas samples from each logical measurement area are analyzed online in real time to obtain the real-time NOx concentration values ​​of each logical measurement area. The real-time NOx concentration values ​​of each logical measurement area are also collected synchronously, timestamped, and their validity is verified.

[0008] In one embodiment, the method of constructing a real-time prediction model for ingress NOx concentration using real-time NOx concentration data measured in different zones and employing a combination of Transformer model and ensemble learning includes: Extract multi-dimensional feature variables that are associated with the real-time NOx concentration data of the partition from the historical operation database. The multi-dimensional feature variables include the real-time measured concentration of each partition, unit load, fuel characteristics, air supply volume, flue gas temperature and oxygen content. Data preprocessing is performed on the extracted multidimensional feature variables, including missing value handling, outlier removal, data smoothing, and standardization; A prediction model with a Transformer encoder as its core structure is constructed, and a self-attention mechanism is used to capture the temporal dependencies and global correlations between feature variables. An ensemble learning strategy is adopted to combine multiple Transformer models initialized with different data subsets or parameters to build a strong predictor; The training set, validation set, and test set are divided using preprocessed historical data. The ensemble prediction model is trained with the goal of minimizing the prediction error, and its hyperparameters are tuned using a validation set. The model performance was evaluated using a test set, resulting in a real-time prediction model for inlet NOx concentration that meets the accuracy requirements.

[0009] In one embodiment, the step of constructing a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration includes: Set target denitrification efficiency and outlet NOx concentration set values; A mathematical model was established to establish the relationship between ammonia injection demand and inlet NOx concentration, flue gas flow rate, and denitrification efficiency. The mathematical model was established based on the principles of chemical reaction kinetics and system design parameters. The predicted inlet NOx concentration, outlet NOx concentration setpoint, and target denitrification efficiency output by the real-time prediction model of inlet NOx concentration are input into the mathematical model to obtain the required predicted total amount of ammonia injection. A feedforward compensation term is introduced to correct the predicted total ammonia injection. The feedforward compensation term is dynamically calculated based on feedback from flue gas temperature changes, catalyst activity decay rate, and ammonia slip rate. The predicted total ammonia injection amount after feedforward correction is used as the setpoint of the total ammonia injection amount control loop to construct the feedforward control algorithm for the total ammonia injection amount prediction.

[0010] In one embodiment, the feedforward control algorithm based on total ammonia injection prediction achieves intelligent and coordinated control and regulation of ammonia injection flow through a three-stage valve system consisting of a total ammonia injection automatic control valve, a zoned ammonia injection automatic regulating valve, and a branch pipe regulating manual valve, which are connected in series. This includes: The ammonia injection total automatic control valve receives the total ammonia demand command from the ammonia injection total prediction feedforward control algorithm and acts as a primary control actuator to macroscopically adjust the total ammonia injection amount. Based on the real-time NOx concentration data measured in each zone, combined with the flue gas flow field distribution, the required ammonia distribution ratio and flow rate setpoint for each zone are calculated. Each zone's corresponding automatic ammonia injection regulating valve receives the required ammonia distribution ratio and flow setpoint for each zone as a secondary control actuator to perform fine ammonia injection flow distribution for each area. The branch pipe regulating hand valves installed on the ammonia injection branch pipes of each level zone serve as a three-level backup and manual fine-tuning mechanism, providing manual intervention capability during maintenance of the automatic control system or abnormal operating conditions to ensure basic ammonia injection function. Establish a collaborative control logic and communication link between three levels of valves. Through the collaborative control logic and communication link, monitor and collect the opening feedback of each level of valve, the actual ammonia injection flow rate, and the denitrification efficiency and NOx concentration changes of the corresponding zone in real time to obtain a closed-loop control feedback data chain. Based on the closed-loop control feedback data, dynamically optimize the zone flow allocation and valve response parameters.

[0011] Furthermore, to achieve the above objectives, the present invention also proposes a flow field optimization device for an ultra-supercritical unit denitrification system, wherein the flow field optimization device is applied to the flow field optimization method for an ultra-supercritical unit denitrification system as described in any one of claims 1 to 6, and the device comprises: A module is built to establish a dynamic flow field model of denitrification flue gas based on the flue structure of the target denitrification system using simulation software. The simulation module is used to determine the location of the flow divider baffle in the trapezoidal variable diameter flue based on the dynamic flow field model and to install it in order to optimize the flue gas flow field distribution in the flue. The data acquisition module is used to perform real-time zoned measurement of NOx concentration in flue gas within the optimized flue using a multi-point mixing method. The construction module is used to construct a real-time prediction model for inlet NOx concentration using real-time NOx concentration data measured in partitions and employing a combination of Transformer model and ensemble learning. The construction module is used to construct a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration. The control module is used to realize the intelligent and coordinated control and regulation of ammonia injection flow through three levels of valves set in series: an automatic control valve for total ammonia injection, a zoned automatic adjustment valve for ammonia injection, and a branch pipe adjustment manual valve, based on the feedforward control algorithm for total ammonia injection prediction.

[0012] In one embodiment, the simulation module is used to perform simulation analysis on the initial flow field in the trapezoidal variable diameter flue based on the dynamic flow field model. The simulation analysis is used to identify key locations where the flow field distribution is uneven, eddies exist, or low-velocity zones exist. Based on the simulation analysis results, various arrangement schemes for the flow dividers are determined. The arrangement schemes include the number of flow dividers, installation positions, installation angles, and geometric parameters. In the dynamic flow field model, each arrangement scheme of the flow divider is loaded and simulation calculations are performed. Comparative analysis of the simulation results of each scheme was conducted, and the evaluation indicators included the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change in system drag. Based on the preset optimization target, a target arrangement scheme is selected from various arrangement schemes of the flow divider baffles. The target arrangement scheme is the flow divider baffle arrangement scheme with the best overall performance. Based on the design parameters of the target layout scheme, a flow divider plate is installed at the actual physical location of the trapezoidal variable diameter flue to optimize the flue gas flow field distribution within the flue.

[0013] Furthermore, to achieve the above objectives, the present invention also proposes a flow field optimization device for an ultra-supercritical unit denitrification system. The flow field optimization device for an ultra-supercritical unit denitrification system includes: a memory, a processor, and an ultra-supercritical unit denitrification system flow field optimization program stored in the memory and executable on the processor. The ultra-supercritical unit denitrification system flow field optimization program is configured to implement the steps of the ultra-supercritical unit denitrification system flow field optimization method as described above.

[0014] Furthermore, to achieve the above objectives, the present invention also proposes a storage medium storing a flow field optimization program for an ultra-supercritical unit denitrification system. When the ultra-supercritical unit denitrification system flow field optimization program is executed by a processor, it implements the steps of the flow field optimization method for an ultra-supercritical unit denitrification system as described above.

[0015] This invention utilizes a dynamic flow field model of denitrification flue gas established based on the flue structure of the target denitrification system to determine and install the appropriate location for the flow divider baffle within the trapezoidal variable-diameter flue, thereby optimizing the flue gas flow field distribution within the flue. Within the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Based on the output of an inlet NOx concentration real-time prediction model constructed using the zoned real-time NOx concentration data, a feedforward control algorithm for predicting total ammonia injection is built. This algorithm, through the series configuration of three valve stages, enables intelligent and coordinated control and regulation of the ammonia injection flow rate. This effectively improves the uniformity of the flue gas flow field distribution within the flue, enhances the accuracy of NOx concentration measurement, and achieves real-time and precise regulation of the ammonia injection rate, thereby improving denitrification efficiency and reducing the risk of ammonia escape. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of the first embodiment of the flow field optimization method for the denitrification system of an ultra-supercritical unit according to the present invention; Figure 2 This is a structural block diagram of the first embodiment of the flow field optimization device for the denitrification system of the ultra-supercritical unit of the present invention.

[0017] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0019] This invention provides a method for optimizing the flow field of a denitrification system in an ultra-supercritical unit, referring to... Figure 1 , Figure 1 This is a schematic flowchart of the first embodiment of the flow field optimization method for a denitrification system of an ultra-supercritical unit according to the present invention.

[0020] In this embodiment, the flow field optimization method for the denitrification system of the ultra-supercritical unit includes the following steps: Step S10: Based on the flue structure of the target denitrification system, establish a dynamic flow field model of the denitrification flue gas using simulation software.

[0021] In this embodiment, the executing entity is a flow field optimization device for the denitrification system of an ultra-supercritical unit. This flow field optimization device has functions such as data processing, data communication, and program execution. The flow field optimization device for the denitrification system of an ultra-supercritical unit can be a computer terminal device or other network device, or other devices with similar functions. This embodiment does not limit the scope of the application.

[0022] It should be noted that with the continuous improvement of environmental standards, the SCR denitrification system of ultra-supercritical coal-fired power plants faces severe challenges in operation. Under deep peak shaving or low-load conditions, the problem of uneven flow field distribution caused by defects in the flue gas structure design is particularly prominent. In particular, the trapezoidal variable-diameter flue gas area is prone to forming eddies and low-velocity zones, resulting in an imbalance in flue gas velocity distribution. This uneven flow field directly affects the mixing effect of flue gas and reducing agent, reducing the reaction rate on the catalyst surface. Insufficient denitrification in some areas and excessive ammonia in others leads to localized ammonia escape and the formation of ammonium bisulfate, further exacerbating the risk of air preheater blockage and equipment corrosion. In existing technologies, flow field optimization largely relies on empirically installed guide vanes, lacking precise positioning and parameter design based on dynamic flow field models, making it difficult to adapt to complex and variable operating conditions. Simultaneously, NOx concentration measurement generally uses single-point sampling methods, failing to capture the regional concentration variation characteristics of the flue gas cross-section, resulting in insufficient basis for ammonia injection control. Ammonia injection systems often employ traditional feedback control strategies, whose response speed lags behind load fluctuations, making it difficult to achieve real-time and precise adjustment of the ammonia injection rate. This leads to fluctuations in denitrification efficiency and increased ammonia slip during load changes. Although some studies have attempted to improve the flow field through structural modifications, they have failed to organically combine flow field optimization, zoned concentration measurement, and intelligent predictive control. The lack of quantitative analysis and coordinated regulation of flue gas dynamics hinders the stable and efficient operation of denitrification systems under complex conditions.

[0023] To address the aforementioned technical issues, this embodiment utilizes a dynamic flow field model of the denitrification flue gas established based on the flue structure of the target denitrification system to determine and install the diameter-separating guide plate within the trapezoidal variable-diameter flue, thereby optimizing the flue gas flow field distribution within the flue. Within the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Based on the output of an inlet NOx concentration real-time prediction model constructed using the zoned real-time NOx concentration data, a feedforward control algorithm for predicting total ammonia injection is built. This algorithm, through the series configuration of three valves, enables intelligent and coordinated control and regulation of the ammonia injection flow rate. This effectively improves the uniformity of the flue gas flow field distribution within the flue, enhances the accuracy of NOx concentration measurement, and achieves real-time and precise regulation of the ammonia injection amount, thereby improving denitrification efficiency and reducing the risk of ammonia escape. Specifically, this can be implemented as follows.

[0024] In practical implementation, the denitrification flue gas dynamic flow field model refers to a model reflecting the flow characteristics of flue gas constructed through numerical simulation methods. In practical applications, this model can be implemented using computational fluid dynamics software, such as ANSYS Fluent or OpenFOAM, primarily to predict the flow field distribution within the flue. The guide vane, specifically the flow divider, is a flow guiding device used to improve the uniformity of the flue gas flow field. Furthermore, its installation location can be determined based on on-site velocity measurement data, for example, by using a thermal anemometer to perform multi-point measurements at key sections of the flue to identify low-velocity areas and determine the installation location. Specifically, the multi-point mixing method refers to a method of measuring NOx concentration in flue gas in different zones. In practical applications, this can be achieved using a fixed probe array combined with manual mixing, for example, mixing flue gas from multiple sampling points through a pipeline and then extracting the mixed sample using a pump to obtain the regional average concentration. As a preferred implementation, the construction of the inlet NOx concentration real-time prediction model can employ a Transformer architecture and an ensemble learning strategy. For example, a single Transformer model can be combined with the Bagging ensemble method to train multiple models using different training subsets and average the output, primarily to capture the temporal dependencies of concentration data. The ammonia injection total quantity prediction feedforward control algorithm refers to a control algorithm that calculates ammonia injection demand based on the output of the prediction model. Furthermore, its construction can be based on empirical formulas, such as using the linear proportional relationship between inlet NOx concentration and ammonia injection quantity for feedforward calculation to achieve advance adjustment of ammonia injection quantity. Specifically, the intelligent collaborative control of ammonia injection flow rate is achieved through a three-stage valve system set in series. In practical applications, PLC system programming can be used to implement sequential control of valve opening, for example, first adjusting the main valve opening and then allocating the proportion of zone valves, thereby achieving the distribution of ammonia injection flow rate. Therefore, the core innovation of this application lies in solving the problems of decreased denitrification efficiency and ammonia escape caused by uneven flow field and NOx distribution deviation during low-load operation by systematically integrating flue gas flow field optimization and intelligent ammonia injection control strategies. The dynamic flow field model established based on the actual flue structure provides a basis for flow field modification. The multi-point mixing method provides a means of measuring NOx concentration. The prediction model combining the Transformer model and ensemble learning realizes the prediction of inlet concentration. The feedforward control algorithm for total ammonia injection prediction, combined with the coordinated control of three-stage valves, enables the ammonia injection flow rate to match changes in flue gas conditions.

[0025] It should be noted that the flow field optimization method for the ultra-supercritical unit denitrification system first establishes a dynamic flow field model of the denitrification flue gas based on the flue geometry of the target denitrification system using commercial fluid dynamics simulation software. This model can accurately simulate the flow characteristics of flue gas in a variable-diameter flue, providing a data foundation for subsequent flow field optimization. Based on the simulation results of this dynamic flow field model, key locations with uneven velocity distribution, eddies, or low-velocity zones within the trapezoidal variable-diameter flue are identified, and the installation positions of the flow divider baffles are determined accordingly. For example, baffles are installed in a specific area 1.5 to 2.0 meters downstream of the flue bend to eliminate local flow defects. After the baffles are installed, the flue gas flow field distribution within the flue is improved. At this point, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Specifically, the optimized flue cross-section is divided into four logical measurement areas, with three NOx concentration sampling probes evenly distributed in each area. The average concentration data of each area is obtained through a hybrid sampling network. Furthermore, utilizing real-time NOx concentration data measured in different zones, combined with multi-dimensional operating parameters such as unit load and flue gas temperature, a real-time inlet NOx concentration prediction model is constructed using a combination of Transformer models and ensemble learning. For example, this model consists of three Transformer encoders based on the BERT architecture, each containing a 12-layer structure. A self-attention mechanism is used to capture the temporal dependencies of concentration changes, and a random forest ensemble strategy is employed to enhance the model's robustness. Based on the output of the prediction model, a feedforward control algorithm for predicting total ammonia injection is constructed. This algorithm calculates the ammonia injection demand based on the principles of chemical reaction kinetics and incorporates dynamic compensation based on flue gas temperature changes and catalyst activity decay parameters. Therefore, the ammonia injection total quantity prediction feedforward control algorithm realizes intelligent collaborative control of ammonia injection flow through three levels of valves set in series: the ammonia injection total quantity automatic control valve, the zone ammonia injection automatic regulating valve, and the branch pipe regulating manual valve. Among them, the ammonia injection total quantity automatic control valve acts as the first-level actuator to regulate the total ammonia quantity, the zone ammonia injection automatic regulating valve acts as the second-level actuator to distribute the regional flow, and the branch pipe regulating manual valve acts as the third-level backup mechanism to provide manual intervention capability. Each level of valve feeds back the opening degree and flow data in real time through the communication link, forming a closed-loop control logic.

[0026] In one embodiment, in the denitrification system of a 600MW ultra-supercritical unit, a dynamic flow field model was established using ANSYS Fluent software. Simulation analysis showed that there was a significant vortex region downstream of the trapezoidal variable-diameter flue. Based on this, a stainless steel guide plate with a thickness of 8 mm was added to the contraction section of the flue cross-section. After optimization, the flue cross-section was divided into four sector-shaped logical measurement areas, and three high-temperature NOx concentration sampling probes were installed in each area. The sampling signals were mixed and analyzed in real time by the continuous emission monitoring system. A real-time prediction model for inlet NOx concentration was trained using historical operating data. This model integrated three Transformer sub-models, and the inputs included characteristic variables such as zone concentration, unit load, and flue gas temperature. In the ammonia injection control link, the automatic control valve for total ammonia injection was an electric regulating valve, the automatic regulating valve for zone ammonia injection was a pneumatic actuator, and the branch pipe regulating manual valve was a stainless steel ball valve. The three-stage valves worked together to execute the predictive feedforward command to ensure dynamic matching between the ammonia injection flow rate and the flue gas flow field and NOx distribution. The above methods effectively solve the problem of decreased denitrification efficiency caused by uneven flue gas flow field distribution and spatial differences in NOx concentration during low-load operation. Specifically, the addition of baffles guided by the dynamic flow field model significantly improves the uniformity of flue gas velocity, avoiding incomplete reaction caused by local low-velocity zones; multi-point mixing method for zonal measurement provides accurate regional concentration distribution data, overcoming the deficiency of insufficient representativeness of a single measuring point; the real-time inlet NOx concentration prediction model combined with feedforward control algorithm adjusts the ammonia injection amount in advance, eliminating the lag of traditional feedback control; the intelligent collaborative mechanism of the three-stage valves achieves precise matching between the total ammonia injection amount and regional distribution, thereby suppressing ammonia escape caused by excessive or insufficient ammonia injection. As a result, the operational stability and environmental performance of the denitrification system are improved, ensuring the continuous achievement of denitrification efficiency standards under variable load conditions.

[0027] In practical implementation, the flow field optimization method of the denitrification system of ultra-supercritical units determines the location of the flow divider baffles in the trapezoidal variable diameter flue based on the dynamic flow field model and installs them to optimize the flue gas flow field distribution in the flue. However, in this process, due to the lack of refined simulation verification and multi-dimensional evaluation mechanism for the baffle arrangement scheme, the selection of parameters such as the installation position, angle and number of baffles depends on experience judgment, which cannot accurately eliminate the uneven flow field distribution, the presence of eddies or low-speed zones, and thus cause the flue gas velocity to fluctuate more and the system resistance to increase abnormally, ultimately affecting the reliability of NOx concentration zoning measurement and the stability of denitrification control.

[0028] Step S20: Based on the dynamic flow field model, determine the location for installing the flow divider baffle in the trapezoidal variable diameter flue and install it to optimize the flue gas flow field distribution in the flue.

[0029] In specific implementation, the location for installing diameter-separating guide vanes within the trapezoidal variable-diameter flue is determined based on the dynamic flow field model, and the vanes are installed to optimize the flue gas flow field distribution within the flue. This further includes: performing simulation analysis on the initial flow field within the trapezoidal variable-diameter flue based on the dynamic flow field model; the simulation analysis is used to identify key locations where the flow field distribution is uneven, vortices exist, or low-velocity regions are present; determining various arrangement schemes for the diameter-separating guide vanes based on the simulation analysis results; the arrangement schemes include the number of guide vanes, installation positions, installation angles, and geometric parameters; and adding guide vanes to the dynamic flow field model. For each arrangement scheme of the flow divider baffle, simulation calculations are performed; the simulation results of each scheme are compared and analyzed, and the evaluation indicators include the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change of system resistance; according to the preset optimization target, a target arrangement scheme is selected from various flow divider baffle arrangement schemes, which is the flow divider baffle arrangement scheme with the best comprehensive performance; according to the design parameters of the target arrangement scheme, flow divider baffles are installed at the actual corresponding physical location of the trapezoidal variable diameter flue to optimize the flue gas field distribution in the flue.

[0030] In practical implementation, the initial flow field of the trapezoidal variable-diameter flue is simulated and analyzed based on a dynamic flow field model to accurately locate key locations where the flow field distribution is uneven, vortices exist, or low-velocity regions exist. Subsequently, based on the analysis results, multiple flow deflector arrangement schemes are generated, covering the combined design of the number of deflectors, installation positions, installation angles, and geometric parameters. Each scheme is loaded sequentially into the dynamic flow field model for simulation calculation to obtain flow field response data. By comparing and analyzing evaluation indicators such as flue gas velocity distribution uniformity, relative standard deviation of velocity at key sections, turbulent kinetic energy distribution, and system resistance changes of each scheme, a multi-dimensional performance evaluation system is formed. Based on preset optimization objectives, the target arrangement scheme with the best comprehensive performance is selected. Finally, the scheme parameters are mapped to the actual physical location of the flue and the deflectors are installed. This process constructs a closed-loop optimization mechanism from flow field defect identification and scheme virtual verification to engineering implementation, ensuring that the selection of deflector parameters takes into account both flow field uniformity improvement and system resistance control, avoiding local optimization deviations caused by empirical judgment.

[0031] It should be noted that the dynamic flow field model refers to a numerical calculation framework used to simulate the dynamic flow behavior of flue gas in the flue of a denitrification system. It can be constructed using computational fluid dynamics software, with the aim of accurately reproducing the transient evolution process of the flue gas flow field. The flow path divider is a flow guide structure installed inside the flue to divide the airflow channel and guide the direction of the flue gas flow. It can be made of high-temperature resistant alloy plates, with the aim of eliminating local flow field defects through physical intervention. The geometric shape parameters in the layout scheme can include the curvature characteristics, plate inclination angle and edge transition form of the guide plate, with the aim of adapting to the flue geometry to optimize the airflow distribution. The relative standard deviation of the velocity at the key section in the evaluation index refers to the ratio of the flue gas velocity fluctuation at a specific section to the average velocity, with the aim of objectively quantifying the level of flow field uniformity.

[0032] In one embodiment, the specific implementation of the scheme of this application is as follows: the flow divider baffle adopts an arc-shaped plate structure and is installed in the starting area of ​​the contraction section of the trapezoidal variable diameter flue, and is fixed to the inner wall of the flue through flange connection; in the simulation analysis, a significant vortex region is identified on the inner side of the 90-degree bend of the flue, and three arrangement schemes are designed accordingly: Scheme 1 uses a single baffle arranged vertically on the inner side of the bend, Scheme 2 uses two baffles symmetrically and inclinedly installed on both sides of the bend, and Scheme 3 uses three baffles radially distributed in the bend area; after dynamic flow field model loading simulation and multi-dimensional evaluation, Scheme 2 performs best in terms of reducing the relative standard deviation of the velocity at the key section and controlling the increase in system resistance, so Scheme 2 is selected for actual installation. The inclination angle of the baffle is dynamically adjusted according to the shape of the flue section to achieve a smooth airflow transition. The above scheme can systematically eliminate the uneven flow field distribution, vortices or low-speed areas in the trapezoidal variable diameter flue, effectively suppress flue gas velocity fluctuations and control abnormal increases in system resistance, thereby providing a stable and uniform flow field environment for real-time measurement of NOx concentration by zone, and ensuring the reliability and stability of the denitrification control process.

[0033] Step S30: In the optimized flue, the NOx concentration in the flue gas is measured in real time by zone using a multi-point mixing method.

[0034] In its specific implementation, the method of using a multi-point mixing approach to perform real-time zoned measurement of NOx concentration in flue gas further includes: dividing the flue cross-section into multiple logical measurement regions based on the optimized flow field characteristics of the flue; arranging multiple NOx concentration sampling probes within each logical measurement region to form a distributed sampling network; mixing the flue gas from each probe within each logical measurement region proportionally through the distributed sampling network to obtain a mixed flue gas sample, which represents the average concentration of the corresponding logical measurement region; and using a continuous emission monitoring system or a rapid analyzer to perform real-time online analysis of the mixed flue gas samples from each logical measurement region to obtain the real-time NOx concentration values ​​of each zone's logical measurement region, and simultaneously collecting, aligning, and validating the real-time NOx concentration values ​​of each logical measurement region.

[0035] In this embodiment, logical measurement areas are scientifically divided based on optimized flue gas flow field characteristics to ensure that the partitions match the actual dynamic characteristics of the flow field. Secondly, a distributed sampling network is deployed in each logical measurement area to obtain comprehensive concentration distribution information through multi-point coverage. Thirdly, the flue gas from each probe in the area is mixed proportionally to generate a representative mixed flue gas sample. Finally, the mixed sample is analyzed online in real time using a continuous emission monitoring system or a rapid analyzer, and synchronous acquisition, timestamp alignment, and validity verification are implemented. This systematically solves the measurement distortion problem caused by uneven NOx concentration distribution in flue gas, providing high-precision and timely input data for subsequent prediction models.

[0036] It should be noted that the logical measurement region refers to the sub-region of flue gas cross-section divided according to the characteristics of the flue gas flow field. Its boundaries can be determined by methods based on the uniformity of velocity distribution or the significance of concentration gradient, with the aim of accurately matching the dynamic characteristics of the flow field to capture spatial differences in concentration. The distributed sampling network refers to a sampling system composed of multiple NOx concentration sampling probes, which can be implemented by equidistant grid arrangement or concentrated arrangement for vortex regions, with the aim of eliminating the influence of random fluctuations in single-point measurements through multi-point coverage. In practical applications, proportional mixing refers to the process of mixing flue gas according to the sampling point location and flow velocity weight, which can be implemented by Venturi mixers or static mixers, with the aim of making the mixed sample truly reflect the average concentration level of the region.

[0037] In one embodiment, the flue gas duct cross-section is divided into four sector-shaped logical measurement areas, with three NOx concentration sampling probes installed in each area. The probes are made of high-temperature resistant stainless steel and connected to the flue gas duct via flanges. A static mixer is used to achieve proportional mixing of the flue gas during the mixing process. A continuous emission monitoring system is employed to perform real-time online analysis of the mixed samples. This technical solution accurately obtains the NOx concentration distribution in each area of ​​the flue gas duct, avoiding data distortion caused by single-point measurements. This improves the input data quality of the inlet NOx concentration prediction model, ensures the accuracy of ammonia injection control, effectively reduces the risk of ammonia escape, and enhances denitrification efficiency. Step S40: Using real-time NOx concentration data measured in the partition, construct a real-time prediction model for the inlet NOx concentration by combining the Transformer model with ensemble learning.

[0038] In its specific implementation, the method of constructing a real-time NOx concentration prediction model for the inlet using real-time NOx concentration data measured by different zones and employing a combination of Transformer models and ensemble learning further includes: extracting multi-dimensional feature variables associated with the real-time NOx concentration data measured by different zones from a historical operating database; these multi-dimensional feature variables include the real-time measured concentration of each zone, unit load, fuel characteristics, air supply volume, flue gas temperature, and oxygen content; performing data preprocessing on the extracted multi-dimensional feature variables, including missing value handling, outlier removal, data smoothing, and standardization; constructing a prediction model with a Transformer encoder as its core structure, utilizing a self-attention mechanism to capture the temporal dependencies and global correlations between feature variables; employing an ensemble learning strategy to combine multiple Transformer models initialized based on different data subsets or parameters to construct a strong predictor; dividing the preprocessed historical data into training, validation, and test sets; training the ensemble prediction model with the goal of minimizing prediction error, and performing hyperparameter tuning using the validation set; and evaluating the model performance using the test set to obtain a real-time NOx concentration prediction model for the inlet that meets the accuracy requirements.

[0039] In this embodiment, the proposed solution first extracts multi-dimensional feature variables from a historical operational database. These variables collectively constitute a complete feature set reflecting the system's operational status, providing a comprehensive data foundation for prediction. Next, data preprocessing is performed on the feature variables to eliminate noise interference and standardize data scale, ensuring the quality of model input. Based on this, a prediction model centered on a Transformer encoder is constructed, utilizing a self-attention mechanism to dynamically calculate the correlation weights between feature variables, effectively capturing the nonlinear temporal characteristics under varying load conditions. Furthermore, an ensemble learning strategy is employed to combine multiple Transformer models, forming a strong predictor through weighted fusion of model outputs, improving prediction stability. Subsequently, the preprocessed data is rigorously divided into training, validation, and test sets. The training process optimizes model parameters with the goal of minimizing prediction error, and the validation set dynamically adjusts hyperparameters to prevent overfitting. Finally, an independent test set is used to evaluate model performance, ensuring that the prediction results accurately reflect actual operating conditions. The entire process, through the organic combination of feature engineering, model architecture innovation, and training strategy optimization, achieves accurate and dynamic prediction of inlet NOx concentration.

[0040] It should be noted that multidimensional feature variables refer to a set of multiple related parameters used to characterize the system's operating state. These can include parameters such as real-time measured concentrations in each zone, unit load, fuel characteristics, air supply volume, flue gas temperature, and oxygen content. The purpose of introducing these parameters is to comprehensively capture internal and external factors affecting the inlet NOx concentration and avoid prediction bias caused by a single data source. Data preprocessing refers to the process of cleaning and transforming the raw data. This can be achieved by using linear interpolation to handle missing values, removing outliers based on the 3σ criterion, applying sliding windows for data smoothing, and standardizing through normalization methods. The aim is to eliminate noise and incompleteness in the actual operating data and ensure the reliability of the input data. The Transformer encoder refers to a deep learning architecture based on a self-attention mechanism. It can be implemented by stacking multi-layer self-attention modules and feedforward neural networks. The purpose of introducing this parameter is to efficiently identify long-distance temporal dependencies between feature variables. Ensemble learning strategies refer to methods that combine multiple basic models to improve overall performance. This can be achieved by using a Bagging approach to integrate Transformer models based on different training subsets. The aim is to reduce prediction variance through model diversity and enhance the system's robustness under complex operating conditions.

[0041] In one embodiment, the historical operation database can specifically be the unit's historical operation logs stored in an industrial-grade time-series database; during multi-dimensional feature variable extraction, real-time concentration data for each zone is obtained from a distributed sampling network, and unit load parameters are read in real-time through the DCS system interface; the Transformer model can be implemented using an open-source deep learning framework, where the self-attention mechanism is configured as a multi-head attention structure; in the ensemble learning strategy, multiple Transformer models initialized with different random seeds are trained, and the prediction results are combined using a weighted average method; data partitioning adopts a time-series segmentation method to ensure the temporal continuity of the training set and the test set; hyperparameter tuning is performed using a Bayesian optimization algorithm, iteratively adjusting the learning rate and network depth with the goal of minimizing the validation set loss. This approach effectively improves the accuracy and real-time response capability of inlet NOx concentration prediction, adapting to scenarios of flue gas flow field fluctuations and rapid changes in operating parameters under varying unit load conditions, thereby providing a reliable basis for ammonia injection control and ensuring the stability of the denitrification system during dynamic operation and the continuous compliance of environmental protection indicators.

[0042] Step S50: Construct a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration.

[0043] In a specific implementation, the step of constructing a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time inlet NOx concentration prediction model further includes: setting a target denitrification efficiency and an outlet NOx concentration setpoint; establishing a mathematical model between the ammonia injection demand and the inlet NOx concentration, flue gas flow rate, and denitrification efficiency, wherein the mathematical model is established based on the principles of chemical reaction kinetics and system design parameters; inputting the predicted inlet NOx concentration, outlet NOx concentration setpoint, and target denitrification efficiency output by the real-time inlet NOx concentration prediction model into the mathematical model to obtain the required predicted total ammonia injection; introducing a feedforward compensation term to correct the predicted total ammonia injection, wherein the feedforward compensation term is dynamically calculated based on feedback from flue gas temperature changes, catalyst activity decay rate, and ammonia slip rate; and using the feedforward-corrected predicted total ammonia injection as the setpoint of the total ammonia injection control loop to construct the feedforward control algorithm for predicting total ammonia injection.

[0044] In this embodiment, the proposed solution provides a clear environmental benchmark for ammonia injection control by setting a target denitrification efficiency and an outlet NOx concentration setpoint. A mathematical model based on chemical reaction kinetics is used as input to calculate the predicted total ammonia injection amount, taking the predicted inlet NOx concentration, outlet NOx concentration setpoint, and target denitrification efficiency output from the real-time inlet NOx concentration prediction model as input. Subsequently, a feedforward compensation term is introduced based on dynamic feedback calculations of flue gas temperature changes, catalyst activity decay rate, and ammonia slip rate to correct the predicted total ammonia injection amount in real time. Finally, the feedforward-corrected predicted total ammonia injection amount is used as the setpoint for the total ammonia injection amount control loop, driving subsequent valve actuators to achieve precise control of the ammonia injection flow rate.

[0045] It should be noted that the mathematical model refers to the ammonia injection demand calculation model based on the principle of chemical reaction kinetics. It can be implemented using an analytical model or a data-driven model. Its purpose is to accurately characterize the intrinsic relationship between the ammonia injection amount and key operating parameters. The feedforward compensation term can be understood as a correction factor for dynamically adjusting the ammonia injection amount. It can be dynamically calculated based on flue gas temperature sensor data, catalyst activity monitoring data, and ammonia slip rate feedback data. Its purpose is to correct and predict the total ammonia injection amount in real time to adapt to dynamic operating factors.

[0046] In one embodiment, the solution of this application is implemented as follows: the mathematical model is specifically implemented as a MATLAB-based calculation module, which integrates chemical reaction kinetic equations; the calculation of the feedforward compensation term specifically uses a weighted average of real-time flue gas temperature data and the output of the catalyst activity decay model to dynamically adjust the correction coefficient. Through the above technical solution, the ammonia injection control accuracy is improved, effectively suppressing ammonia escape or denitrification efficiency fluctuations caused by dynamic factors under variable load conditions, ensuring stable and efficient operation of the denitrification system.

[0047] Step S60: Based on the feedforward control algorithm for total ammonia injection prediction, the three-stage valves of total ammonia injection automatic control valve, zoned ammonia injection automatic regulating valve, and branch pipe regulating manual valve are set in series to realize the intelligent collaborative control and regulation of ammonia injection flow.

[0048] In its specific implementation, the ammonia injection total quantity prediction feedforward control algorithm achieves intelligent and coordinated control and regulation of ammonia injection flow through a three-stage valve system consisting of a total ammonia injection total quantity automatic control valve, a zone ammonia injection automatic regulating valve, and a branch pipe regulating manual valve. This further includes: the total ammonia injection total quantity automatic control valve receiving the total ammonia demand command from the ammonia injection total quantity prediction feedforward control algorithm, acting as a primary control actuator to macroscopically regulate the total ammonia injection quantity; calculating the required ammonia distribution ratio and flow rate setpoint for each zone based on real-time measured NOx concentration data for each zone and the flue gas flow field distribution; and the zone ammonia injection automatic regulating valve receiving the required ammonia distribution ratio and flow rate setpoint for each zone. The flow setpoint serves as a secondary control actuator for precise ammonia injection flow allocation across different zones. The branch pipe regulating valves, installed on the ammonia injection branch pipes at each zone level, act as a tertiary backup and manual fine-tuning mechanism, providing manual intervention capabilities during automatic control system maintenance or abnormal operating conditions to ensure basic ammonia injection functionality. A collaborative control logic and communication link are established between the three levels of valves. This collaborative control logic and communication link allows for real-time monitoring and acquisition of valve opening feedback, actual ammonia injection flow, and changes in denitrification efficiency and NOx concentration in the corresponding zones, resulting in a closed-loop control feedback data chain. Based on this closed-loop control feedback data, the zone flow allocation and valve response parameters are dynamically optimized.

[0049] In this embodiment, the total ammonia demand command received by the ammonia injection total quantity prediction feedforward control algorithm is macroscopically adjusted through the ammonia injection total quantity automatic control valve to ensure that the total ammonia injection quantity is synchronized with the predicted demand. At the same time, the required ammonia distribution ratio and flow setpoint for each zone are calculated based on the real-time measured NOx concentration data of each zone and the flue gas flow field distribution, and the zone ammonia injection automatic regulating valve performs fine distribution. The branch pipe regulating manual valve provides manual intervention capability as a third-level backup mechanism. By establishing a collaborative control logic and communication link, the valve opening, ammonia injection flow rate and denitrification effect are monitored in real time to form a closed-loop control feedback data chain. Based on this data chain, the zone flow distribution and valve response parameters are dynamically optimized.

[0050] It should be noted that the automatic ammonia injection total control valve refers to the actuator used for macroscopic adjustment of the total ammonia injection volume. It can be implemented using an electric or pneumatic regulating valve, with the aim of ensuring that the total ammonia injection volume is strictly synchronized with the system's predicted demand. The automatic ammonia injection control valve for zoned injection refers to the actuator used for fine-tuning the ammonia injection flow rate in each zone. It can be implemented using a proportional-integral-derivative control valve, with the aim of achieving precise adaptation of the ammonia injection flow rate. The branch pipe regulating manual valve refers to the manual valve installed on the branch pipes of ammonia injection at each level of zone. It can be implemented using a gate valve or a stop valve, with the aim of providing manual intervention capability during maintenance of the automatic control system or abnormal operating conditions. The collaborative control logic and communication link refers to the system used to realize data interaction and control between the three levels of valves. It can be implemented using industrial Ethernet or fieldbus communication protocols, with the aim of forming a closed-loop control feedback data chain.

[0051] In one embodiment, the automatic control valve for total ammonia injection is an electric butterfly valve, the automatic regulating valve for zoned ammonia injection is a pneumatic regulating valve, and the manual regulating valve for branch pipes is a manual shut-off valve. The coordinated control logic is implemented through a programmable logic controller and communicates via industrial Ethernet. This solution achieves precise allocation of ammonia injection flow, effectively reducing excessive or insufficient ammonia injection caused by uneven NOx concentration distribution within the flue, thereby stabilizing denitrification efficiency and reducing the risk of ammonia escape.

[0052] In this embodiment, a dynamic flow field model of denitrification flue gas based on the flue structure of the target denitrification system is used to determine the location and install the flow divider baffle in the trapezoidal variable diameter flue to optimize the flue gas flow field distribution. In the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Based on the output of the inlet NOx concentration real-time prediction model constructed using the zoned real-time NOx concentration data, a feedforward control algorithm for predicting total ammonia injection is built. This algorithm, through the series setting of three valves, enables intelligent and coordinated control and regulation of the ammonia injection flow rate. This effectively improves the uniformity of the flue gas flow field distribution, enhances the accuracy of NOx concentration measurement, and achieves real-time and precise regulation of the ammonia injection amount, thereby improving denitrification efficiency and reducing the risk of ammonia escape.

[0053] Furthermore, this embodiment of the invention also proposes a storage medium storing a flow field optimization program for an ultra-supercritical unit denitrification system. When the ultra-supercritical unit denitrification system flow field optimization program is executed by a processor, it implements the steps of the flow field optimization method for an ultra-supercritical unit denitrification system as described above.

[0054] Reference Figure 2 , Figure 2 This is a structural block diagram of the first embodiment of the flow field optimization device for the denitrification system of the ultra-supercritical unit of the present invention.

[0055] like Figure 2As shown, the flow field optimization device for the denitrification system of an ultra-supercritical unit proposed in this embodiment of the invention includes: Module 10 is used to build a dynamic flow field model of denitrification flue gas based on the flue structure of the target denitrification system using simulation software; Simulation module 20 is used to determine the location of the flow divider baffle in the trapezoidal variable diameter flue based on the dynamic flow field model and to install it in order to optimize the flue gas flow field distribution in the flue. The acquisition module 30 is used to perform real-time zoned measurement of NOx concentration in flue gas within the optimized flue using a multi-point mixing method. The construction module 10 is used to construct a real-time prediction model for inlet NOx concentration using real-time NOx concentration data measured in partitions and employing a combination of Transformer model and ensemble learning. The construction module 10 is used to construct a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration. The control module 40 is used to realize the intelligent and coordinated control and regulation of ammonia injection flow through three valves set in series: an automatic control valve for total ammonia injection, a zoned automatic adjustment valve for ammonia injection, and a branch pipe adjustment manual valve, based on the feedforward control algorithm for total ammonia injection prediction.

[0056] In this embodiment, a dynamic flow field model of denitrification flue gas based on the flue structure of the target denitrification system is used to determine the location and install the flow divider baffle in the trapezoidal variable diameter flue to optimize the flue gas flow field distribution. In the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Based on the output of the inlet NOx concentration real-time prediction model constructed using the zoned real-time NOx concentration data, a feedforward control algorithm for predicting total ammonia injection is built. This algorithm, through the series setting of three valves, enables intelligent and coordinated control and regulation of the ammonia injection flow rate. This effectively improves the uniformity of the flue gas flow field distribution, enhances the accuracy of NOx concentration measurement, and achieves real-time and precise regulation of the ammonia injection amount, thereby improving denitrification efficiency and reducing the risk of ammonia escape.

[0057] In one embodiment, the simulation module 10 is used to perform simulation analysis on the initial flow field in the trapezoidal variable diameter flue based on the dynamic flow field model. The simulation analysis is used to identify key locations where the flow field distribution is uneven, eddies exist, or low-speed zones exist. Based on the simulation analysis results, various arrangement schemes for the flow dividers are determined. The arrangement schemes include the number of flow dividers, installation positions, installation angles, and geometric parameters. In the dynamic flow field model, each arrangement scheme of the flow divider is loaded and simulation calculations are performed. Comparative analysis of the simulation results of each scheme was conducted, and the evaluation indicators included the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change in system drag. Based on the preset optimization target, a target arrangement scheme is selected from various arrangement schemes of the flow divider baffles. The target arrangement scheme is the flow divider baffle arrangement scheme with the best overall performance. Based on the design parameters of the target layout scheme, a flow divider plate is installed at the actual physical location of the trapezoidal variable diameter flue to optimize the flue gas flow field distribution within the flue.

[0058] This application embodiment also provides a flow field optimization device for an ultra-supercritical unit denitrification system, including a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other through the communication bus. The memory is used to store the flow field optimization program for the ultra-supercritical unit denitrification system. When the processor executes the program stored in the memory, it implements the above-mentioned flow field optimization method for the ultra-supercritical unit denitrification system.

[0059] The communication bus mentioned in the flow field optimization equipment of the ultra-supercritical unit denitrification system can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc.

[0060] The communication interface is used for communication between the flow field optimization equipment of the denitrification system of the above-mentioned ultra-supercritical unit and other equipment.

[0061] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0062] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0063] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0064] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0065] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0066] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0067] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0068] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0069] In addition, for technical details not described in detail in this embodiment, please refer to the flow field optimization method of the denitrification system of ultra-supercritical unit provided in any embodiment of the present invention, which will not be repeated here.

[0070] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0071] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0072] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0073] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

[0074] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

Claims

1. A method for optimizing the flow field of a denitrification system in an ultra-supercritical unit, characterized in that, The flow field optimization method for the denitrification system of the ultra-supercritical unit includes: A dynamic flow field model of denitrification flue gas was established using simulation software based on the flue structure of the target denitrification system. Based on the dynamic flow field model, the location for installing the flow divider baffle in the trapezoidal variable diameter flue was determined and the baffle baffle was installed to optimize the flue gas flow field distribution in the flue. Within the optimized flue, a multi-point mixing method is used to perform real-time zoned measurement of NOx concentration in the flue gas. Using real-time NOx concentration data measured in different zones, a real-time prediction model for ingress NOx concentration is constructed by combining the Transformer model with ensemble learning. A feedforward control algorithm for predicting total ammonia injection volume is constructed based on the output of the real-time prediction model of the inlet NOx concentration. Based on the feedforward control algorithm for total ammonia injection prediction, a three-stage valve system consisting of a total ammonia injection automatic control valve, a zoned ammonia injection automatic regulating valve, and a branch pipe regulating manual valve is connected in series to achieve intelligent and coordinated control and regulation of ammonia injection flow.

2. The flow field optimization method for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The step of determining and installing the diameter-separating guide plate within the trapezoidal variable-diameter flue based on the dynamic flow field model to optimize the flue gas flow field distribution within the flue includes: Based on the dynamic flow field model, the initial flow field in the trapezoidal variable diameter flue is simulated and analyzed. The simulation analysis is used to identify key locations where the flow field distribution is uneven, vortices exist, or low-speed zones are present. Based on the simulation analysis results, various arrangement schemes for the flow dividers are determined. The arrangement schemes include the number of flow dividers, installation positions, installation angles, and geometric parameters. In the dynamic flow field model, each arrangement scheme of the flow divider is loaded and simulation calculations are performed. Comparative analysis of the simulation results of each scheme was conducted, and the evaluation indicators included the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change in system drag. Based on the preset optimization target, a target arrangement scheme is selected from various arrangement schemes of the flow divider baffles. The target arrangement scheme is the flow divider baffle arrangement scheme with the best overall performance. Based on the design parameters of the target layout scheme, a flow divider plate is installed at the actual physical location of the trapezoidal variable diameter flue to optimize the flue gas flow field distribution within the flue.

3. The flow field optimization method for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The method of using a multi-point mixing approach to perform real-time zoned measurement of NOx concentration in flue gas includes: Based on the optimized flow field characteristics of the flue, the cross-section of the flue is divided into multiple logical measurement regions; Multiple NOx concentration sampling probes are deployed within each logical measurement area to form a distributed sampling network; The distributed sampling network mixes the flue gas from each probe within each logical measurement area proportionally to obtain a mixed flue gas sample, which is used to represent the average concentration of the corresponding logical measurement area. Using a continuous emission monitoring system or rapid analyzer, the mixed flue gas samples from each logical measurement area are analyzed online in real time to obtain the real-time NOx concentration values ​​of each logical measurement area. The real-time NOx concentration values ​​of each logical measurement area are also collected synchronously, timestamped, and their validity is verified.

4. The flow field optimization method for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The method of constructing a real-time prediction model for ingress NOx concentration using real-time NOx concentration data measured in different zones, employing a combination of Transformer model and ensemble learning, includes: Extract multi-dimensional feature variables that are associated with the real-time NOx concentration data of the partition from the historical operation database. The multi-dimensional feature variables include the real-time measured concentration of each partition, unit load, fuel characteristics, air supply volume, flue gas temperature and oxygen content. Data preprocessing is performed on the extracted multidimensional feature variables, including missing value handling, outlier removal, data smoothing, and standardization; A prediction model with a Transformer encoder as its core structure is constructed, and a self-attention mechanism is used to capture the temporal dependencies and global correlations between feature variables. An ensemble learning strategy is adopted to combine multiple Transformer models initialized with different data subsets or parameters to build a strong predictor; The training set, validation set, and test set are divided using preprocessed historical data. The ensemble prediction model is trained with the goal of minimizing the prediction error, and its hyperparameters are tuned using a validation set. The model performance was evaluated using a test set, resulting in a real-time prediction model for inlet NOx concentration that meets the accuracy requirements.

5. The flow field optimization method for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration includes: Set target denitrification efficiency and outlet NOx concentration set values; A mathematical model was established to establish the relationship between ammonia injection demand and inlet NOx concentration, flue gas flow rate, and denitrification efficiency. The mathematical model was established based on the principles of chemical reaction kinetics and system design parameters. The predicted inlet NOx concentration, outlet NOx concentration setpoint, and target denitrification efficiency output by the real-time prediction model of inlet NOx concentration are input into the mathematical model to obtain the required predicted total amount of ammonia injection. A feedforward compensation term is introduced to correct the predicted total ammonia injection. The feedforward compensation term is dynamically calculated based on feedback from flue gas temperature changes, catalyst activity decay rate, and ammonia slip rate. The predicted total ammonia injection amount after feedforward correction is used as the setpoint of the total ammonia injection amount control loop to construct the feedforward control algorithm for the total ammonia injection amount prediction.

6. The flow field optimization method for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The feedforward control algorithm based on total ammonia injection prediction achieves intelligent and coordinated control and regulation of ammonia injection flow through a three-stage valve system consisting of a total ammonia injection automatic control valve, a zoned ammonia injection automatic regulating valve, and a branch pipe regulating manual valve. This includes: The ammonia injection total automatic control valve receives the total ammonia demand command from the ammonia injection total prediction feedforward control algorithm and acts as a primary control actuator to macroscopically adjust the total ammonia injection amount. Based on the real-time NOx concentration data measured in each zone, combined with the flue gas flow field distribution, the required ammonia distribution ratio and flow rate setpoint for each zone are calculated. Each zone's corresponding automatic ammonia injection regulating valve receives the required ammonia distribution ratio and flow setpoint for each zone as a secondary control actuator to perform fine ammonia injection flow distribution for each area. The branch pipe regulating hand valves installed on the ammonia injection branch pipes of each level zone serve as a three-level backup and manual fine-tuning mechanism, providing manual intervention capability during maintenance of the automatic control system or abnormal operating conditions to ensure basic ammonia injection function. Establish a collaborative control logic and communication link between three levels of valves. Through the collaborative control logic and communication link, monitor and collect the opening feedback of each level of valve, the actual ammonia injection flow rate, and the denitrification efficiency and NOx concentration changes of the corresponding zone in real time to obtain a closed-loop control feedback data chain. Based on the closed-loop control feedback data, dynamically optimize the zone flow allocation and valve response parameters.

7. A flow field optimization device for a denitrification system of an ultra-supercritical unit, characterized in that, The flow field optimization device for the denitrification system of an ultra-supercritical unit is applied to the flow field optimization method for the denitrification system of an ultra-supercritical unit as described in any one of claims 1 to 6. The device comprises: A module is built to establish a dynamic flow field model of denitrification flue gas based on the flue structure of the target denitrification system using simulation software. The simulation module is used to determine the location of the flow divider baffle in the trapezoidal variable diameter flue based on the dynamic flow field model and to install it in order to optimize the flue gas flow field distribution in the flue. The data acquisition module is used to perform real-time zoned measurement of NOx concentration in flue gas within the optimized flue using a multi-point mixing method. The construction module is used to construct a real-time prediction model for inlet NOx concentration using real-time NOx concentration data measured in partitions and employing a combination of Transformer model and ensemble learning. The construction module is used to construct a feedforward control algorithm for predicting total ammonia injection based on the output of the real-time prediction model of the inlet NOx concentration. The control module is used to realize the intelligent and coordinated control and regulation of ammonia injection flow through three levels of valves set in series: an automatic control valve for total ammonia injection, a zoned automatic adjustment valve for ammonia injection, and a branch pipe adjustment manual valve, based on the feedforward control algorithm for total ammonia injection prediction.

8. The flow field optimization device for the denitrification system of an ultra-supercritical unit as described in claim 1, characterized in that, The simulation module is used to perform simulation analysis on the initial flow field in the trapezoidal variable diameter flue based on the dynamic flow field model. The simulation analysis is used to identify key locations where the flow field distribution is uneven, vortices exist, or low-speed zones are present. Based on the simulation analysis results, various arrangement schemes for the flow dividers are determined. The arrangement schemes include the number of flow dividers, installation positions, installation angles, and geometric parameters. In the dynamic flow field model, each arrangement scheme of the flow divider is loaded and simulation calculations are performed. Comparative analysis of the simulation results of each scheme was conducted, and the evaluation indicators included the uniformity of flue gas velocity distribution, the relative standard deviation of velocity at key sections, the distribution of turbulent kinetic energy, and the change in system drag. Based on the preset optimization target, a target arrangement scheme is selected from various arrangement schemes of the flow divider baffles. The target arrangement scheme is the flow divider baffle arrangement scheme with the best overall performance. Based on the design parameters of the target layout scheme, a flow divider plate is installed at the actual physical location of the trapezoidal variable diameter flue to optimize the flue gas flow field distribution within the flue.

9. A flow field optimization device for a denitrification system of an ultra-supercritical unit, characterized in that, The flow field optimization device for the ultra-supercritical unit denitrification system includes: a memory, a processor, and an ultra-supercritical unit denitrification system flow field optimization program stored in the memory and executable on the processor. The ultra-supercritical unit denitrification system flow field optimization program is configured to implement the steps of the ultra-supercritical unit denitrification system flow field optimization method as described in any one of claims 1 to 6.

10. A storage medium, characterized in that, The storage medium stores a flow field optimization program for the denitrification system of an ultra-supercritical unit. When the ultra-supercritical unit denitrification system flow field optimization program is executed by the processor, it implements the steps of the flow field optimization method for the ultra-supercritical unit denitrification system as described in any one of claims 1 to 6.

Images