Precise ammonia injection control method, device and system for scr denitration

Abstract

The application discloses a precise ammonia injection control method, device and system for SCR denitration, the precise ammonia injection control method is provided with two-stage ammonia injection grids and matched mother pipes at the inlet of an SCR reactor, at least one mother pipe is provided with a staged adjusting valve, the ammonia injection grid with partition adjusting capacity is provided with multiple partitions, and each partition is matched with a partition adjusting valve. The control process adopts hierarchical control logic: the inlet NOx concentration is predicted by using a deep learning model, and the total ammonia injection amount is calculated in combination with the flue gas flow; the two-stage mother pipe flow is dynamically distributed through the staged adjusting valve; the adjustment amount is generated through zero-sum constraint and normalization according to the NOx deviation at the partition outlet, and the precise fine adjustment of the partition is completed. The application further adds a PID correction and a whole-process feedback loop, adopts a double-layer closed loop algorithm and a recursive least square method to correct the valve characteristics, and limits the ammonia injection proportion of a single-stage grid. The application solves the problems of measurement lag, flow ratio rigidity and uneven partition ammonia injection of a traditional scheme, can effectively improve the denitration precision and reduce ammonia escape.

Description

Technical Field

[0001] This invention relates to the field of flue gas denitrification technology in coal-fired power plants, and in particular to a precise ammonia injection control method, device, and system for SCR denitrification. Background Technology

[0002] Selective catalytic reduction (SCR) denitrification technology is widely used due to its high denitrification efficiency and stable operation. However, in order to control ammonia escape, reduce ammonia consumption and protect the catalyst while ensuring high denitrification efficiency, it is necessary to achieve high precision and uniform distribution of ammonia injection amount and injection position.

[0003] Based on the above practical needs, improving the uniformity and control accuracy of ammonia injection distribution has become a key engineering problem in SCR system optimization, and has promoted the proposal and research of various precise control schemes for zoned ammonia injection. In the prior art, utility model patent CN213966009U discloses a precise ammonia injection zoned control system for denitrification based on multi-point online synchronous measurement. This system divides the ammonia injection grid into multiple ammonia injection zones, each with an independent ammonia injection device, and installs online NOx detection instruments at the outlet of each zone. Based on the NOx concentration at the outlet of each zone, combined with a fuzzy PID algorithm, the ammonia injection valves of each zone are adjusted to achieve precise ammonia injection distribution. Invention patent CN109603525A proposes a denitrification zoned ammonia injection control method based on NOx concentration non-uniformity judgment. This method monitors the NOx distribution at the SCR reactor outlet in real time, calculates its non-uniformity (CV), and dynamically adjusts the zoned ammonia supply valves and ammonia injection branch valves based on the number of ammonia injection branches, the NOx concentration in each zone, and the operating conditions of the coal mill, thereby improving the online adjustment accuracy. The invention patent with publication number CN108380043A discloses a method for regulating and controlling the ammonia injection in a zoned SCR denitrification device. This method adopts a single-layer ammonia injection pipeline structure, with only a single-stage ammonia injection main pipe and multiple zoned branch pipes downstream of the main pipe. Each branch pipe is equipped with a zoned leveling valve. By monitoring important parameters such as the NOx concentration at the inlet of the SCR denitrification device, the NOx concentration in the zoned outlet, the total amount of ammonia injected, and the ammonia flow rate in each zone, the measured value of the ammonia flow rate in each zone is used as the closed-loop signal for regulating and controlling the ammonia injection zoned leveling valve. Combined with the NOx concentration in the inlet and outlet zones of the SCR reactor, the target ammonia flow rate in each zone is calculated. By adjusting the zoned leveling valve, the uniformity of ammonia injection is improved, which alleviates to some extent the problems of poor regulation quality and high ammonia slip caused by the lag in NOx measurement at a single measuring point.

[0004] While existing technologies have achieved some success in improving ammonia injection distribution and control accuracy, limitations remain at the engineering implementation level. The main reason is that, to meet the needs of large-diameter flow supply and operational reliability under wide operating conditions of SCR units, most field ammonia injection grids often use a single large-stroke master valve or large-diameter regulating valve as the main control actuator. Although these valves can achieve significant regulation, their valve-flow characteristics are nonlinear, have dead zones and hysteresis, and their valve resolution and control bandwidth are limited. Therefore, their ability to perform fine, small-amplitude, and rapid zonal regulation is insufficient, making it difficult to meet the demand for minute-level, rapid-response zonal ammonia injection. On the other hand, modifying field systems usually involves large-scale pipeline mechanical alterations, resulting in high costs and hindering large-scale engineering implementation. Furthermore, existing control strategies often logically conflate the calculation of total ammonia injection with zonal allocation or control each zone independently, lacking global conservation constraints. This leads to "competition" between zones or total volume drift, making it difficult to balance overall and local control objectives. Summary of the Invention

[0005] A brief overview of embodiments of the invention is provided below to provide a basic understanding of certain aspects of the invention. It should be understood that this overview is not an exhaustive summary of the invention. It is not intended to identify key or essential parts of the invention, nor is it intended to limit the scope of the invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to the more detailed description that follows.

[0006] To address the technical shortcomings of existing SCR denitrification two-stage ammonia injection systems, such as lagging ammonia injection calculation, rigid two-stage flow ratio, uneven ammonia injection in different zones, and inability to self-correct drift of regulating valve characteristics, this invention proposes a two-stage linkage precision ammonia injection control method for SCR denitrification. This method relies on deep learning to predict the inlet NOx concentration in advance to realize pre-calculation of ammonia injection, dynamically allocates the flow of the two-stage ammonia injection main pipe, and performs zoned balance fine-tuning to improve denitrification efficiency and reduce ammonia escape.

[0007] According to a first aspect of this application, a precise ammonia injection control method for SCR denitrification is provided, applied to a zoned ammonia injection system for SCR denitrification. A first-stage and a second-stage ammonia injection grid are arranged along the flue gas flow direction in the flue gas inlet duct of the SCR reactor. Both stages of the ammonia injection grid are equipped with a main pipe, and zoned branch pipes are provided below the main pipe. The sum of the flow rates of the two main pipes is the total ammonia injection amount. At least one stage of the ammonia injection grid has a graded regulating valve on its corresponding main pipe to distribute the two-stage flow rates. The ammonia injection grid with zoned regulation capability is configured with multiple zones, each zone being equipped with a zoned regulating valve. The precise ammonia injection control method includes: Inlet NOx Prediction: Predict the NOx concentration at the flue gas inlet; this step can be performed using a pre-trained deep learning model, or other prediction methods such as machine learning and data fitting. Total ammonia injection calculation: Calculate the total amount of ammonia to be injected by combining the predicted inlet NOx concentration and flue gas flow rate; Two-stage main pipe linkage distribution: Based on the total ammonia injection volume, the valve position of the staged regulating valve is adjusted through closed-loop iteration to dynamically distribute the flow rates of the first-stage main pipe and the second-stage main pipe; the sum of the flow rates of the two-stage main pipes is always equal to the total ammonia injection volume; Zoned ammonia injection fine-tuning control: Collect the measured NOx concentration at the outlet of each zone, compare it with the preset target NOx concentration to calculate the NOx concentration deviation of each zone, generate the relative adjustment amount of ammonia injection for each zone through zero-sum constraint and normalization, and convert the relative adjustment amount of each zone into the valve position adjustment command of the regulating valve of each zone to complete the precise fine-tuning of the zone.

[0008] As a feasible solution, the first-stage and second-stage ammonia injection grids are respectively newly added and existing ammonia injection grids, with the two stages of grids spaced apart along the flue gas flow direction. By employing a two-stage ammonia injection strategy, the denitrification system can respond quickly and correct for small-range ammonia quantity deviations with high precision. Through a hierarchical control strategy, the hysteresis of the denitrification system is effectively reduced, achieving high-precision, rapid, and robust coordinated control of zoned ammonia injection while meeting total quantity and safety constraints. This effectively reduces the non-uniformity of NOx at the outlet section and suppresses ammonia escape.

[0009] Furthermore, the total ammonia injection calculation step also includes a total ammonia injection correction process: the total amount of NOx to be reduced is calculated by combining the predicted NOx concentration and the real-time flue gas flow rate, and the baseline total ammonia injection amount is obtained; the deviation between the measured NOx value at the reactor outlet and the NOx target value of the zoned ammonia injection system is sent to the PID controller to generate a correction amount, and the correction amount is added to the baseline total ammonia injection amount to obtain the final total ammonia injection amount. Feedback loops are configured between each step: the measured NOx value at the reactor outlet after precise fine-tuning by zone is fed back to the total ammonia injection calculation step to participate in the correction of the next cycle, forming a closed-loop control of the system.

[0010] Furthermore, the two-stage busbar linkage allocation step is implemented using a dual-layer closed-loop iterative fine-tuning algorithm: the upper-level controller (DCS) performs a total allocation calculation and generates the target flow rate for each busbar periodically (e.g., every 2 to 10 seconds), while the lower-level actuator (PLC) performs closed-loop valve position control based on flow feedback periodically (e.g., every 0.1 to 0.5 seconds). The upper-level controller summarizes the flow rate and valve position data over the long period, uses the recursive least squares method to identify the valve-flow sensitivity online, and issues the valve position correction amount for the next period accordingly. This process is iterated until the busbar flow rate approaches the target flow rate. For example, the control target for the busbar flow rate deviation is that the deviation between the actual flow rate and the target flow rate is less than ±1.0%.

[0011] Furthermore, in the zoned ammonia injection fine-tuning control step, the zone flow weight is determined based on the flue gas flow rate of each zone, and the unnormalized adjustment amount of each zone is calculated in combination with the zone flow weight. Then, the unnormalized adjustment amount of all zones is subjected to mean-reduction processing so that the sum of the adjustment amounts of each zone is zero to satisfy the zero-sum constraint and complete the normalization operation. The relative adjustment amount after mean-reduction processing is converted into a valve position command through a mapping coefficient and issued, and the fine-tuning is completed through the zoned regulating valve.

[0012] Furthermore, a deep learning model is used for prediction in the NOx prediction step. As a preferred embodiment, the deep learning model is preferably a Seq2Seq model, and the model input parameters include at least boiler load, furnace temperature, total air volume, furnace negative pressure, total fuel quantity, coal mill feed rate, secondary damper opening, and flue gas oxygen content.

[0013] Furthermore, the main pipe corresponding to the zonal regulating valve handles 20% to 40% of the total ammonia injection volume.

[0014] According to a second aspect of this application, a precision ammonia injection control device for SCR denitrification is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when running the computer program, implements the aforementioned precision ammonia injection control method.

[0015] According to a third aspect of this application, an SCR denitrification zoned ammonia injection system is provided, the system including the aforementioned precision ammonia injection control device.

[0016] According to a fourth aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described precise ammonia injection control method.

[0017] This application employs a two-stage ammonia injection strategy, enabling the denitrification system to respond rapidly and correct for small-range ammonia quantity deviations with high precision. By adopting a hierarchical control strategy, it effectively reduces the hysteresis of the denitrification system, achieving high-precision, rapid, and robust coordinated control of zoned ammonia injection while meeting total quantity and safety constraints. This effectively reduces the non-uniformity of NOx at the outlet section and suppresses ammonia escape. Compared with existing technologies, the advantages of this application are: 1) This invention proposes dividing the ammonia injection system into a basic support layer (the original ammonia injection grid) and a fine-tuning operation layer (a newly added ammonia injection grid). The basic support layer continues to handle the majority of the ammonia supply (60%–80% of the total ammonia injection volume) to ensure the reliability of the ammonia supply under various operating conditions. The fine-tuning operation layer handles 20%–40% of the total ammonia injection volume and is equipped with zone regulating valves, responsible for rapid and precise zone ammonia injection compensation. Compared with existing technologies that rely solely on single-stage grids or large-stroke valves, this significantly improves the accuracy and response speed of correcting local ammonia concentration anomalies without weakening the stability of the main ammonia supply. This functional decoupling of coarse adjustment to ensure stability in the basic layer and fine adjustment to correct deviations in the fine-tuning layer is not simply adding a grid, but rather achieving dynamic redistribution of the two-stage flow within a preset range from the control architecture perspective, significantly improving the accuracy and response speed of correcting local ammonia concentration anomalies. In the existing SCR denitrification field, traditional ammonia injection systems generally adopt a design scheme of single-stage bar screen combined with large-stroke regulating valve. Due to the nonlinearity, dead zone and hysteresis of valve-flow characteristics, such valves have insufficient fine adjustment capability. At the same time, existing control strategies often confuse the calculation of total ammonia injection with the allocation of zones, lacking global conservation constraints, resulting in inter-zone competition for quantity or total quantity drift.

[0018] 2) By adopting a hierarchical control strategy, the hysteresis of the denitrification system is reduced, and the transient response capability is improved, thereby effectively reducing the non-uniformity of NOx at the outlet section. Specifically, this invention uses a Seq2Seq deep learning model in the inlet NOx prediction stage to make real-time predictions based on parameters such as boiler load, furnace temperature, total air volume, and flue gas oxygen content, achieving feedforward control 2 minutes in advance. This shortens the effective response lag of the system to sudden operating conditions by approximately 30% to 60%. This prediction, combined with feedback composite control logic, achieves advanced adjustment in the time dimension, significantly improving the transient response capability. 3) A dual-layer closed-loop iterative fine-tuning algorithm is adopted, consisting of an upper-layer DCS (2-10 second cycle) and a lower-layer PLC (0.1-0.5 second cycle). Short-cycle PLC execution with small step sizes is introduced to meet transient response requirements. Meanwhile, the DCS uses recursive least squares (RLS) to online identify valve-flow sensitivity over a longer period to compensate for valve nonlinearity and hysteresis, ensuring that the deviation between the actual flow rate and the target flow rate in the main pipe is less than ±1.0%. This effectively improves ammonia injection execution accuracy, reduces steady-state and transient errors, and enhances the system's robustness against valve aging, blockage, and sudden changes in operating conditions. The root mean square error of the main pipe flow tracking can be reduced by 50%. This adaptive control algorithm significantly improves ammonia injection execution accuracy and system robustness. 4) The proposed zoned ammonia injection fine-tuning method calculates the unnormalized adjustment amount based on the NOx concentration deviation at the outlet of each zone and the flue gas flow weight of each zone. By removing the mean, the sum of the adjustment amounts for each zone is made zero, and complementary adjustments are made to each zone, eliminating local ammonia concentration deviations without increasing the overall ammonia injection amount. This mechanism fundamentally avoids the problem of inter-zone contention or total volume drift, allowing for simultaneous consideration of zone uniformity and total volume conservation. This discrete two-stage decoupled architecture breaks through the conventional thinking that ammonia injection must be centrally controlled by a single device, enabling physical separation of large-flow coarse adjustment and small-flow fine adjustment, executed by different levels. This design concept, which achieves reliability and accuracy in layers, is not a simple expansion of a single-stage grid, but a redefinition of the functional architecture of the ammonia injection system.

[0019] In summary, this application can reduce ammonia consumption and the risk of ammonia escape while ensuring denitrification effect, and also takes into account environmental protection and operating costs, thus having good engineering applicability and economy. Attached Figure Description

[0020] The present invention can be better understood by referring to the description given below in conjunction with the accompanying drawings, in which the same or similar reference numerals are used throughout the drawings to denote the same or similar parts. These drawings, together with the following detailed description, are incorporated in and form part of this specification, and are used to further illustrate preferred embodiments of the invention and explain the principles and advantages of the invention. In the drawings: Figure 1 This is a schematic diagram of the SCR denitrification zoned ammonia injection system in Embodiment 1 of the present invention; Figure 2 This is a flowchart illustrating the hierarchical control strategy used in Embodiment 1 of the present invention. Figure 3 This is a schematic diagram of the ammonia injection zone in Embodiment 2 of the present invention; Figure 4 This is a visual comparison diagram of the predicted and true values ​​of the Seq2Seq model on the test set in Embodiment 2 of the present invention. Figure 5a and Figure 5b The diagrams show the NOx non-uniformity curves at the SCR outlet sections on sides A and B before and after using the zoned ammonia injection fine-tuning method of the present invention. Detailed Implementation

[0021] Embodiments of the present invention will now be described with reference to the accompanying drawings. Elements and features described in one drawing or embodiment of the invention may be combined with elements and features shown in one or more other drawings or embodiments. It should be noted that, for clarity, representations and descriptions of components and processes unrelated to the present invention and known to those skilled in the art have been omitted from the drawings and description.

[0022] Example 1 This embodiment provides a two-stage linkage precision ammonia injection control method for SCR denitrification, which combines a two-stage ammonia injection grid architecture with a hierarchical control strategy to achieve coordinated operation of structural hardware and control strategy.

[0023] Figure 1 This is a schematic diagram of the SCR denitrification zoned ammonia injection system in this embodiment. This embodiment adds an auxiliary ammonia injection grid to the existing denitrification ammonia injection system to construct a graded control structure: along the flue gas flow direction, the upstream is the first-stage grid (Add-AIG, fine-tuning operation layer), and the downstream is the second-stage grid (Exist-AIG, basic protection layer), forming a two-stage ammonia injection structure.

[0024] The second-stage bar screen serves as the basic support layer, retaining the original layout and providing a stable basic ammonia injection flow rate with uniformly arranged nozzles. It does not have zoning control valves on its body, ensuring reliable ammonia supply under all operating conditions. The first-stage bar screen serves as the fine-tuning operation layer, with multiple zones, each equipped with a zoning control valve (or branch pipe control valve). The installation position of the zoning control valves is not restricted, allowing for flexible adjustment of the additional ammonia injection rate in each zone, achieving high-precision fine-tuning. Both stages of the bar screen are equipped with independent main pipes, each with a zoning control valve (or main valve) for distributing the two flow rates. The sum of the flow rates from both main pipes remains constant as the total ammonia injection rate of the system.

[0025] This system divides ammonia injection zones according to the actual dimensions of the flue. A single flue is divided into several equally wide zones along its depth, typically between 4 and 8 zones. The first and second stage grids, hereinafter referred to as Add-AIG and Exist-AIG, can have the same or different number of zones depending on the actual site conditions. Exist-AIG is located in the vertical flue at the inlet of the denitrification reactor, maintaining its existing layout. Add-AIG is also located in the vertical flue at the inlet of the denitrification reactor, upstream of Exist-AIG. To ensure sufficient mixing of the ammonia gas injected from Add-AIG with the flue gas before reaching Exist-AIG, the two stages of ammonia injection grids are spaced a certain distance apart, ensuring good uniformity of the ammonia gas before it finally enters the catalyst. The axial spacing between the two-stage grids is preferably set to 3 to 5 meters. This spacing takes into account the flue gas velocity, turbulence intensity and ammonia diffusion characteristics, ensuring that the ammonia gas injected from Add-AIG is fully mixed before entering the catalyst layer, improving the uniformity of distribution and denitrification efficiency, and reducing the risk of ammonia escape.

[0026] The ammonia injection volume is proportionally distributed between the two-stage grids: Add-AIG accounts for 20%-40% of the total ammonia injection volume, and this proportion can be dynamically adjusted within the range; Exist-AIG handles the remaining 60%-80% of the ammonia injection volume. This ammonia injection volume distribution ensures that the base layer provides the main ammonia supply, guaranteeing reliable ammonia supply under all operating conditions; while the ammonia volume of Add-AIG is fine-tuned through zone valves to achieve precise compensation for each zone.

[0027] This embodiment organically combines a two-stage ammonia injection strategy with control functions through a hierarchical control strategy. By clearly defining hierarchical responsibilities and information flow, it achieves coordinated regulation of the total ammonia injection volume, main pipe allocation, and zoned fine-tuning. This minimizes NOx non-uniformity at the outlet cross-section and suppresses ammonia escape while ensuring accurate total volume control. The hierarchical control strategy in this embodiment comprises four main components: inlet NOx prediction, total ammonia injection control, two-stage linkage control, and zoned ammonia injection fine-tuning. The complete flowchart is shown below. Figure 2 As shown, the specific execution steps are as follows: Step 1, Inlet NOx Prediction: Historical and real-time operating parameters such as boiler load, air volume ratio, coal mill configuration, flue gas flow, oxygen distribution, and temperature distribution are collected through the DCS system. These parameters are then combined with the measured NOx concentration at the SCR inlet to train a deep learning model. During operation, the trained model is used to predict the NOx concentration at the SCR flue gas inlet in real time, overcoming the lag in field measurements and rapidly responding to changes in boiler operating conditions. The model's prediction results serve as feedforward input to guide the total ammonia injection control.

[0028] Step 2, total ammonia injection control: This is achieved by combining feedforward control with PID feedback regulation, and a full-process feedback loop is configured.

[0029] First, based on the predicted inlet NOx concentration and real-time flue gas flow rate, the total amount of NOx to be reduced is calculated. This total NOx is then converted into the required ammonia injection rate according to the ammonia-nitrogen molar ratio, and recorded as the baseline total ammonia injection rate. Subsequently, the measured NOx value at the reactor outlet is collected, and its deviation from the system's target NOx value is calculated (target value). The measured value is used to input the deviation into the PID controller to obtain the feedback correction value. Finally, the feedback correction value is superimposed on the reference total ammonia injection value to obtain the final total ammonia injection value. This process corrects the total ammonia injection control signal, realizing closed-loop adjustment of the total ammonia injection valve.

[0030] The system is set up with a full-process feedback loop: after the zone fine-tuning is completed, the measured NOx value at the reactor outlet is fed back to this step to participate in the total correction of the next cycle, forming a closed-loop control.

[0031] Step 3, Two-stage header linkage control: High-precision flow meters are configured on the ammonia injection headers of Add-AIG and Exist-AIG respectively. Flow rate is used as the closed-loop control variable. A dual-layer closed-loop iterative fine-tuning algorithm with a long cycle for the upper-level controller and a short cycle for the lower-level actuator is used to achieve dynamic flow distribution and valve characteristic correction. The upper-level controller can be implemented using a DCS, which determines the initial distribution ratio based on the total ammonia injection volume and the current operating conditions, and generates the target flow rate for each header. The lower-level actuator can be implemented using a PLC. The PLC implements closed-loop flow control for the corresponding header valves, using flow meter feedback as the controlled variable in the control loop, and valve position as the actuated variable, which is fed back for online identification and correction of the valve-flow mapping relationship.

[0032] Considering the characteristics of ammonia injection valves such as low sensitivity, nonlinearity, and hysteresis, the upper-level controller DCS has an execution cycle of 2 to 10 seconds, periodically completing the total allocation calculation and generating the target flow of the two-stage main pipe; the lower-level actuator PLC has an execution cycle of 0.1 to 0.5 seconds, performing valve position adjustment in a small step manner to complete local closed-loop valve position control based on flow feedback, while also performing valve position rate limiting, upper and lower limit constraints, and local interlock protection.

[0033] The DCS aggregates historical data on main pipe flow and stage control valve positions over a long period. It uses the recursive least squares (RLS) method to identify valve-flow sensitivity online. Based on the identification results, it generates stage control valve position correction values ​​and sends them to the PLC. The PLC performs valve position correction in small steps (0.5% / cycle) in the next short cycle, iterating and adjusting until the main pipe flow approaches the target value (the deviation between the actual main pipe flow and the target flow is less than ±1.0%), thus completing the dynamic distribution of the two-stage main pipe flow and the online correction of valve nonlinearity, hysteresis, and characteristic drift.

[0034] The sum of the main pipe flow rates of Add-AIG and Exist-AIG is always equal to the total ammonia injection rate, and the distribution ratio of Add-AIG fluctuates dynamically within a preset range (default 30%, adjustable within the set upper and lower limits) to achieve smooth and reliable two-stage linkage control while meeting the outlet NOx target and reducing cross-sectional non-uniformity.

[0035] Step 4, Zoned Ammonia Injection Fine-tuning Control: This step achieves fine-tuning only through the zoned regulating valves of Add-AIG. During the adjustment process, the total ammonia injection amount of Add-AIG remains constant. The problem of uneven NOx distribution in the zones is eliminated by using flow weight, zero-sum constraint and normalization algorithm: By calculating the deviation between the NOx at the outlet of each zone and the target NOx at the system outlet, the relative adjustment amount of each zone is generated, and this relative adjustment amount is mapped to the valve position command of Add-AIG for each zone, thereby minimizing the non-uniformity of NOx at the outlet cross-section.

[0036] The fine-tuning of ammonia injection in this zone includes the following steps: (1) Data acquisition and weight calculation: Assume that the total number of system partitions is n, and collect the measured NOx concentration at the outlet of the i-th partition (i=1, 2, ..., n) in sequence. Zoned flue gas flow Calculate the partition traffic weight corresponding to the i-th partition. , ; (2) Deviation calculation: Calculate the NOx concentration deviation Δ for each zone. That is, the measured NOx concentration at the outlet of this zone. Difference from the system's preset NOx target concentration: Δ = ; in The target NOx concentration is preset for the system.

[0037] (3) Calculation of unnormalized adjustment: The unnormalized adjustment of each partition is calculated by combining the partition weight and concentration deviation. : ; in The optional prediction feedforward term, i.e., the predicted total ammonia injection amount, is obtained by the inlet NOx prediction model; For proportional gain, This is the feedforward gain, which can be initially set. Once the prediction model stabilizes, the size can be gradually increased. .

[0038] (4) Zero-sum constraint and normalization: To ensure that the total ammonia content of Add-AIG does not change, the unnormalized adjustment amount of all partitions is subjected to mean removal processing so that the sum of the adjustment amounts of each partition is zero, satisfying the zero-sum constraint. The specific calculation formula is as follows: ; This represents the relative adjustment amount of the i-th partition after zero-sum constraint and normalization processing.

[0039] (5) Valve position mapping and execution: The relative adjustment amount after removing the mean is mapped to the valve position command of the zone control valve through a mapping coefficient: ; in For the first The baseline valve position of the zoned control valve in the Add-AIG is the reference opening that should be issued according to the initial allocation under given operating conditions and total quantity. This indicates that the final distribution will be sent to the [number]th [number]. The command valve position of the zone regulating valve. These are mapping coefficients, used to convert dimensionless... Mapped to valve opening, It can be a preset constant, or the flow gain of the valve / actuator can be used as... The value. Finally, the valve position command for each partition. The data is sent from the DCS to the corresponding PLC, which then performs lower-level opening control, rate limiting, and local interlock detection.

[0040] Example 2 This embodiment selects a 1000MW coal-fired power plant SCR denitrification unit as the application object. This SCR denitrification unit adopts a double-row parallel structure, divided into reactors on both sides, A and B; the flue cross-section of a single reactor is 3.5m (width) × 19.65m (length). See also... Figure 3 In this embodiment, the ammonia injection section on one side is evenly divided into 6 equal-width zones along the depth of the flue. Each equal-width zone is equipped with 4 ammonia injection branch pipes, for a total of 24 ammonia injection branch pipes on one side. NOx sampling points are set downstream of each zone to collect zone concentration feedback signals.

[0041] In the vertical flue at the reactor inlet on each side, two stages of ammonia injection grids are arranged sequentially along the flue gas flow direction. The upstream stage is a newly added fine-tuning grid (first-stage grid, Add-AIG), and the downstream stage is the existing basic grid (second-stage grid, Exist-AIG). The axial distance between the two stages of grids is approximately 5m. Under normal operating conditions, Exist-AIG handles approximately 60%–80% of the total ammonia injection. A uniform array of nozzles is arranged at the end of each ammonia injection branch pipe, and no zone regulating valves are installed to ensure the reliability of the basic ammonia supply. Add-AIG handles the remaining approximately 20%–40% of the total ammonia injection. An electric regulating valve is installed before the branch pipe of each zone, equipped with a valve position feedback device for zone fine-tuning and rapid response.

[0042] This embodiment uses a Seq2Seq model to predict inlet NOx concentration. The model input parameters include boiler load, furnace temperature, total air volume, furnace negative pressure, total fuel quantity, coal mill feed rate, secondary damper opening, and flue gas oxygen content. The model training dataset consists of field operation data from 00:00 on April 28, 2023 to 12:13 on June 12, 2023, with a sampling frequency of 1 minute, accumulating 65,534 valid data points. A time-series partitioning method is used, dividing the dataset according to time order, with the first 80% of the samples used as the training set and the last 20% as the test set.

[0043] Model test set performance metrics: Root Mean Square Error (RMSE) = 1.97, Coefficient of Determination (R²) 2 =0.985, Mean Absolute Error (MAE) = 0.338. Visual comparison of predicted and true values ​​of the Seq2Seq model on the test set: Figure 4 As shown, the predicted curve and the measured curve highly overlap, indicating high overall tracking accuracy and demonstrating the model's excellent fitting ability. When the above predictive model is used for feedforward control, the response lag of the denitrification system can be significantly reduced. When a 2-minute prediction lead is selected for control decision-making, experimental and engineering applications show that the effective response lag to sudden operating conditions can be shortened by approximately 30%–60%.

[0044] To achieve dynamic linkage and optimized distribution between the two-stage ammonia injection grids, high-precision flow meters are installed on the ammonia injection headers of Add-AIG and Exist-AIG, and a two-layer control architecture is adopted. In one specific implementation, the upper-level controller DCS has an execution cycle of 5 seconds. Combining the predicted inlet NOx value and the measured outlet NOx deviation, it calculates the total ammonia injection volume and issues the target flow rate for both headers. The DCS then sends the target flow rate to the lower-level PLC actuator via Modbus. The lower-level PLC actuator has a sampling and execution cycle of 0.1 seconds. Within each sampling cycle, it performs closed-loop control based on flow feedback, adjusting the valve position of the corresponding header valve to ensure that the measured flow rate tracks the target flow rate.

[0045] To compensate for low valve sensitivity, nonlinearity, and hysteresis, this embodiment employs a dual-layer closed-loop iterative fine-tuning algorithm: The DCS summarizes the 50 most recently generated flow-valve position data points from the PLC every 5 seconds and estimates the current valve-flow sensitivity using recursive least squares (RLS); based on the sensitivity update value obtained from RLS, the DCS calculates and issues a correction amount for the target valve position. In one specific implementation, the correction cycle is 5 seconds, and the PLC performs the correction in small steps over several subsequent 0.1-second cycles, transmitting the results in real time. In one specific implementation, the small step size is 0.5% of the valve position change amplitude per cycle; the short-cycle closed loop and long-cycle identification iterate alternately until the main pipe flow deviation is less than a preset percentage of the target flow, for example, the deviation between the actual main pipe flow and the target flow is less than ±1.0%.

[0046] Regarding the allocation strategy of Add-AIG and Exist-AIG, the DCS defaults to setting Add-AIG to 30% of the total ammonia injection, with Exist-AIG accounting for the remaining 70%. During operation, the DCS allows the proportion of Add-AIG to dynamically fluctuate within a preset upper and lower limit of 20%–40% to respond to local cross-sectional NOx deviations. However, all allocation adjustments are made without changing the total ammonia injection to avoid blindly increasing the overall ammonia injection. In this embodiment, after implementing the above two-level linkage control strategy, the main pipe flow tracking error is significantly reduced. The root mean square error (RMSE) of the main pipe flow tracking is reduced from approximately 1.6 kg / h before the modification to approximately 0.8 kg / h, a decrease of approximately 50%. The consistency between the actual flow and the target flow is greatly improved, and the flow control accuracy is significantly enhanced.

[0047] In this embodiment, the zoned ammonia injection fine-tuning function is achieved only through the first-stage grid Add-AIG. During the adjustment process, the total ammonia injection amount of Add-AIG remains constant, and the complementary adjustment of the ammonia injection amount of each zone is completed by relying on flow weight, zero-sum constraint and normalization algorithm.

[0048] Figure 5a and Figure 5b The figures show the comparison curves of NOx non-uniformity at the outlet sections of reactors A and B before and after optimization. As can be seen from the figures, before optimization, the NOx concentration non-uniformity at the SCR outlet section ranged from 30% to 55%, indicating significant differences in NOx distribution in the flue gas. After optimization using the method of this invention, the non-uniformity was reduced to below 15%, and the uniformity of NOx concentration distribution at the SCR outlet section was significantly improved.

[0049] Field operation data shows that this invention has good ammonia-saving effects under different load conditions: ammonia consumption decreases by about 8%–12% under high load conditions compared to before the modification, and ammonia consumption decreases by about 22%–32% under low load conditions, with a comprehensive average ammonia-saving rate of about 22%. The measured average hourly ammonia consumption reduction is about 45 kg; based on an estimated 4500 hours of annual unit operation, this translates to a reduction of approximately 202.5 tons of ammonia consumption per year. Therefore, this method can significantly reduce the ammonia consumption of the unit while ensuring high-quality denitrification, demonstrating good economic benefits and promotional value.

[0050] Example 3 The present invention also provides a precision ammonia injection control device for SCR denitrification, which can perform the precision ammonia injection control method described in any of the foregoing embodiments.

[0051] The precision ammonia injection control device includes a processor, a memory, and a bus. The memory and processor are interconnected via the bus to enable data and instruction interaction. The memory stores computer programs; the processor comprises one or more processing cores that execute the computer programs stored in the memory to implement the precision ammonia injection control method of any of the above embodiments.

[0052] As an executable solution, the device can be a computer unit, which may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above-described structure of the computer unit is merely an example and does not constitute a limitation on the computer unit. It may include more or fewer components, or combine certain components, or use different components. For example, the computer unit may also include input / output devices, network access devices, buses, etc., and this embodiment of the invention does not limit this.

[0053] As an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, 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, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0054] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0055] Example 4 This invention also provides a zoned ammonia injection system for SCR denitrification. The system includes the aforementioned precision ammonia injection control device and supporting hardware such as two-stage ammonia injection grids, pipeline valves, and data acquisition units. By running a preset computer program through the precision ammonia injection control device, the system completes the entire process control, including inlet NOx prediction, total ammonia injection calculation, two-stage main pipe flow distribution, and zoned ammonia injection fine-tuning, thereby achieving precise SCR flue gas denitrification.

[0056] Example 5 The present invention also provides a computer-readable storage medium storing a computer program; when the computer program is executed by a processor, it can implement the precise ammonia injection control method described in Embodiment 1 or Embodiment 2 above.

[0057] Furthermore, the present invention also provides a computer program product comprising computer program instructions, which, when executed by a processor, perform all the steps of the aforementioned precise ammonia injection control method.

[0058] It should be emphasized that the term "including / comprises" as used herein refers to the presence of a feature, element, step, or component, but does not exclude the presence or addition of one or more other features, elements, steps, or components.

[0059] Furthermore, the method of the present invention is not limited to being executed in the chronological order described in the specification, but may also be executed in other chronological orders, in parallel, or independently. Therefore, the execution order of the method described in this specification does not constitute a limitation on the technical scope of the present invention.

[0060] Although the invention has been disclosed above through the description of specific embodiments, it should be understood that all the embodiments and examples described above are exemplary and not restrictive. Those skilled in the art can design various modifications, improvements, or equivalents to the invention within the spirit and scope of the appended claims. These modifications, improvements, or equivalents should also be considered to be included within the protection scope of the invention.

Claims

1. A precise ammonia injection control method for SCR denitrification, applied to a zoned ammonia injection system for SCR denitrification, characterized in that, In the flue gas inlet duct of the SCR reactor, first-stage and second-stage ammonia injection grids are installed along the flue gas flow direction. Both stages of ammonia injection grids are equipped with a main pipe, and branch pipes are installed below the main pipe. The sum of the flow rates of the two main pipes is the total ammonia injection volume. At least one stage of the ammonia injection grid has a staged regulating valve on its corresponding main pipe to distribute the flow rates between the two stages. The ammonia injection grid with zoned regulation capability is configured with multiple zones, each zone equipped with a zoned regulating valve. The precise ammonia injection control method includes: Inlet NOx Prediction: Predicts the NOx concentration at the flue gas inlet; Total ammonia injection calculation: Calculate the total amount of ammonia to be injected by combining the predicted inlet NOx concentration and flue gas flow rate; Two-stage main pipe linkage distribution: Based on the total ammonia injection volume, the valve position of the staged regulating valve is adjusted through closed-loop iteration to dynamically distribute the flow rate between the first-stage main pipe and the second-stage main pipe. Zoned ammonia injection fine-tuning control: Collect the measured NOx concentration at the outlet of each zone, compare it with the preset target NOx concentration to calculate the NOx concentration deviation of each zone, generate the relative adjustment amount of ammonia injection for each zone through zero-sum constraint and normalization, and convert the relative adjustment amount of each zone into the valve position adjustment command of the regulating valve of each zone to complete the precise fine-tuning of the zone.

2. The precise ammonia injection control method according to claim 1, characterized in that, In the NOx prediction step at the inlet, a pre-trained deep learning model is used to predict the NOx concentration at the flue gas inlet. The deep learning model is a Seq2Seq model, and the model input parameters include at least boiler load, furnace temperature, total air volume, furnace negative pressure, total fuel quantity, coal mill feed rate, secondary damper opening, and flue gas oxygen content.

3. The precise ammonia injection control method according to claim 1, characterized in that, The total ammonia injection calculation step also includes a total ammonia injection correction process: combining the predicted NOx concentration and real-time flue gas flow rate to calculate the total NOx to be reduced, and converting it into a baseline total ammonia injection amount; The deviation between the measured NOx value at the reactor outlet and the target NOx value of the zoned ammonia injection system is sent to the PID controller to generate a correction amount. The correction amount is then added to the baseline total ammonia injection amount to obtain the final total ammonia injection amount. Feedback loops are configured between each step: After precise fine-tuning of the zones, the measured NOx value at the reactor outlet is fed back to the total ammonia injection calculation step to participate in the correction of the next cycle, forming a closed-loop control.

4. The precise ammonia injection control method according to claim 1, characterized in that, The two-stage main pipe linkage distribution step is implemented using a dual-layer closed-loop iterative fine-tuning algorithm: the upper-level controller performs total distribution calculations and generates target flow rates for each main pipe in a long-term cycle, while the lower-level actuator performs closed-loop valve position control based on flow feedback in a short-term cycle; the upper-level controller summarizes the main pipe flow rate and the valve position data of the stage regulating valves in the long-term cycle, uses the recursive least squares method to identify the valve-flow sensitivity online, and issues the valve position correction amount of the stage regulating valves in the next cycle accordingly, iterating cyclically until the main pipe flow rate deviation meets the requirements.

5. The precise ammonia injection control method according to claim 1, characterized in that, The main pipe corresponding to the aforementioned zone regulating valve is provided, and the amount of ammonia injected accounts for 20% to 40% of the total ammonia injection.

6. The precise ammonia injection control method according to claim 4, characterized in that, The control target for the flow deviation of the main pipe is: the deviation between the actual flow and the target flow is less than ±1.0%.

7. The precise ammonia injection control method according to claim 1, characterized in that, In the zoned ammonia injection fine-tuning control step, the zone flow weight is determined based on the flue gas flow of each zone, and the unnormalized adjustment amount of each zone is calculated in combination with the zone flow weight. The unnormalized adjustment amounts of all partitions are processed by removing the mean, so that the sum of the adjustment amounts of each partition is zero to satisfy the zero-sum constraint, thus completing the normalization operation; the relative adjustment amounts after removing the mean are converted into valve position commands through mapping coefficients and issued, and fine-tuning is completed through the partition control valve.

8. A precision ammonia injection control device for SCR denitrification, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor, when running the computer program, implementing the precise ammonia injection control method according to any one of claims 1 to 7.

9. A zoned ammonia injection system for SCR denitrification, characterized in that, Includes the precision ammonia injection control device as described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the precise ammonia injection control method according to any one of claims 1 to 7.

Images