SCR denitration time-delay compensation estimation-feedback compound control method and device based on characteristic flue gas parameters

Abstract

The invention relates to the technical field of automatic flue gas denitration control of thermal generator sets, in particular to an SCR denitration time-lag compensation estimation-feedback composite control method and device based on characteristic flue gas parameters, and the method comprises the steps: collecting historical data of an SCR denitration system, and building a dynamic characteristic model database of the SCR denitration system under different flue gas flows and temperatures; characteristic flue gas parameters are collected in real time, a dynamic characteristic model under the current working condition is generated, a corresponding estimation model is constructed to obtain the estimated outlet NOx concentration, and actually measured outlet NOx is filtered; and a main feedback and pre-estimation compensation composite control framework is constructed, an ammonia spraying valve control instruction is generated through order reduction compensation processing, opening and closing of an ammonia spraying valve are controlled in advance, and the NOx concentration of an outlet is rapidly adjusted. According to the method, the problem of large lag existing in denitration control can be solved, the change of the opening degree of the valve required under the real-time characteristic flue gas condition is predicted, advanced adjustment is achieved, and optimal control over the SCR denitration system is achieved.

Description

Technical Field

[0001] This invention belongs to the field of automatic control technology for flue gas denitrification in thermal power generating units, and relates to ammonia injection regulation of SCR reactor, NOx emission control, and object time-delay inertia modeling and compensation. More specifically, it relates to a composite control method and device for SCR denitrification time-delay compensation prediction-feedback based on characteristic flue gas parameters. Background Technology

[0002] Selective Catalytic Reduction (SCR) denitrification technology is widely used in flue gas treatment of coal-fired power units due to its high removal efficiency for nitrogen oxides (NOx). An SCR system typically includes an ammonia source and delivery system, ammonia injection unit, reactor and catalyst bed, and measurement and control unit. Its operation must balance ammonia consumption, ammonia slip, and ammonium salt deposition risks under emission constraints. The SCR reaction is significantly affected by operating parameters such as flue gas temperature, flow rate, and inlet NOx concentration: temperature determines the catalytic activity range and reaction rate; flow rate affects residence time and mixing and mass transfer conditions; and fluctuations in inlet NOx alter the relationship between theoretical ammonia demand and denitrification efficiency. Simultaneously, catalyst activity decay, ash accumulation and blockage, and changes in the tendency to form ammonium bisulfate can cause time-dependent drift in the target gain and dynamic response.

[0003] At the control level, ammonia injection regulation exhibits significant lag and inertia: ammonia gas experiences transport delays from valve regulation to injection mixing and reaction within the catalyst layer; NOx analyzer sampling and signal filtering introduce measurement delays and noise. During deep peak shaving and rapid load changes, fuel and air distribution variations cause short-term, drastic fluctuations in inlet NOx, flue gas temperature, and flue gas volume, easily leading to outlet NOx overshoot, oscillation, or slow response; increasing ammonia injection margin to avoid exceeding limits may result in increased ammonia escape, increased ammonia consumption, and blockage of downstream heat exchange equipment. Existing projects often employ methods based on distributed control systems (DCS), such as PID feedback, proportional / feedforward correction, segmented tuning, and zoned ammonia injection balancing. These methods are easy to implement, but the parameters are usually tuned under limited operating conditions. They are prone to insufficient robustness and decreased control accuracy when faced with rapid changes in operating conditions, time-varying objects, and measurement noise. Although some model prediction or lag compensation approaches can improve dynamic performance, they have high requirements for model accuracy, online identification stability, and anti-interference processing, and there are still difficulties in engineering adaptation.

[0004] In the prior art, Chinese patent CN110794685A discloses a method based on mismatch compensation Smith predictive control, which optimizes PID and introduces a feedback compensator. However, it heavily relies on a static model obtained from step disturbance experiments, making it difficult to capture the nonlinear effects of dynamic characteristic parameters such as flue gas velocity and temperature on time delay in real time. CN113893685A discloses a method based on hysteresis inertia compensation, which achieves inertia compensation through a state observer. However, the observer has limited robustness to changes in system parameters and fails to fully utilize the characteristic information from the flue gas side to dynamically correct the time delay model. These technical solutions still suffer from severe model mismatch and limited compensation accuracy when facing drastic parameter fluctuations under wide load peak shaving.

[0005] In summary, existing control strategies based on fixed-parameter PID controllers exhibit significant shortcomings when dealing with SCR denitrification systems, which possess strong nonlinearity, time-varying characteristics, and large time delays, especially under varying operating conditions such as peak shaving. These shortcomings include poor adaptability, lag in regulation, and susceptibility to overshoot and oscillation. Therefore, effectively compensating for the time delay of the denitrification process under variable operating environments, while simultaneously ensuring the robustness and dynamic response speed of the control system, is a pressing technical problem to be solved in the field of flue gas denitrification automation. Summary of the Invention

[0006] (a) Purpose of the invention In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a SCR denitrification time delay compensation prediction-feedback composite control method and device based on characteristic flue gas parameters. By characterizing the operating conditions with characteristic parameters such as flue gas temperature and flow rate and constructing a corresponding SCR dynamic characteristic model accordingly, a prediction link is established by combining the pure time delay and measurement delay characteristics of the object and coordinated with feedback regulation to generate ammonia injection valve control commands. This achieves rapid and stable tracking and disturbance suppression of outlet NOx, thereby solving the problem that the existing coal-fired unit SCR denitrification system has a large delay and large inertia in the control of ammonia injection, making it difficult to achieve precise NOx emission control.

[0007] (II) Technical Solution To achieve the objective of this invention and solve its technical problems, the present invention adopts the following technical solution: The first objective of this invention is to provide a composite control method for SCR denitrification time delay compensation prediction-feedback based on characteristic flue gas parameters, comprising at least the following steps: SS1. Model Parameter Database Construction: Collect historical operating data of SCR denitrification system, identify corresponding SCR dynamic characteristic model parameters under different characteristic flue gas parameter conditions, establish the mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters, and form a model parameter database; SS2. Current operating condition model determination: Real-time acquisition of characteristic flue gas parameters of the current operating condition, obtaining SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters based on the model parameter database and the mapping relationship, generating the SCR dynamic characteristic model under the current operating condition and using it as the subsequently controlled object; SS3. Predicted Output Generation: Based on the SCR dynamic characteristic model under the current operating conditions, a prediction model is constructed to predict the outlet NOx response under ammonia injection control to obtain the predicted outlet NOx concentration, and the real-time acquired outlet NOx concentration signal is low-pass filtered to obtain the filtered outlet NOx concentration. SS4. Construction of Prediction-Feedback Composite Control Framework: Construct a prediction-feedback composite control framework, which includes a main feedback control loop and a prediction compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the setpoint of the NOx concentration at the outlet. The prediction compensation control loop generates the prediction compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet through order reduction compensation processing. SS5. Adaptive tuning of controller parameters: Based on the SCR dynamic characteristic model parameters under the current operating conditions, adaptive calculation and tuning of controller parameters in the main feedback control loop and the predictive compensation control loop are performed to ensure that the main feedback control loop meets the stability requirements and achieves closed-loop tracking and disturbance suppression of the outlet NOx concentration, and the predictive compensation control loop achieves time delay compensation and dynamic response optimization, thereby reducing the impact of the controlled object's time delay and operating condition changes on the closed-loop dynamic performance. SS6. Composite control quantity synthesis: The main feedback control quantity and the estimated compensation control quantity are superimposed to form the ammonia injection composite control quantity, and converted into the ammonia injection valve opening control command based on the corresponding relationship; SS7. Control command output and cyclic update: Output the ammonia injection valve opening control command to the ammonia injection valve actuator of the SCR denitrification system, and cyclically execute steps SS2~SS6 according to the preset control cycle to continuously update the controlled object and output the ammonia injection valve opening control command under changing operating conditions.

[0008] The second objective of this invention is to provide a composite control device for SCR denitrification time delay compensation prediction-feedback based on characteristic flue gas parameters. The method employing the aforementioned composite control method for SCR denitrification time delay compensation prediction-feedback based on characteristic flue gas parameters includes: The model parameter database module is used to store the SCR dynamic characteristic model parameters identified from historical operating data under different characteristic flue gas parameter conditions, as well as the mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters. The current operating condition model generation module is used to receive characteristic flue gas parameters in real time, obtain the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters based on the model parameter database and in combination with the mapping relationship, and generate the SCR dynamic characteristic model under the current operating condition as the controlled object. The predicted output generation module is used to build a prediction model based on the SCR dynamic characteristic model under the current operating conditions, to predict the outlet NOx concentration by predicting the outlet NOx response under ammonia injection control, and to perform low-pass filtering on the real-time outlet NOx concentration to obtain the filtered outlet NOx concentration. The composite control framework module includes a main feedback control loop and a predictive compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the set value of the NOx concentration at the outlet. The predictive compensation control loop generates the predictive compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet, combined with the order reduction compensation processing. The parameter adaptive tuning module is used to adaptively calculate and tune the controller parameters in the main feedback control loop and the predictive compensation control loop based on the SCR dynamic characteristic model parameters under the current operating conditions. The composite control quantity synthesis module is used to superimpose the main feedback control quantity and the estimated compensation control quantity to form the ammonia injection composite control quantity, and convert it into the ammonia injection valve opening control command. The cyclic update module is used to drive each module to execute cyclically according to a preset control cycle, so as to continuously update the controlled object and continuously output the ammonia injection valve opening control command under changing operating conditions.

[0009] (III) Technical Effects Compared with the prior art, the SCR denitrification time delay compensation prediction-feedback composite control method and device based on characteristic flue gas parameters provided by the present invention has the following beneficial and significant technical effects: (1) This invention collects historical operating data of the SCR denitrification system, identifies the corresponding SCR dynamic characteristic model parameters under different characteristic flue gas parameter conditions, establishes a database of SCR dynamic characteristic models under different flue gas flow rates and temperatures, constructs the current operating condition SCR dynamic characteristic model based on the real-time changes of characteristic flue gas parameters, and calculates the ammonia injection demand under characteristic flue gas parameters by constructing a prediction model and a prediction-feedback composite control framework. This enables advance control of the opening and closing of the ammonia injection valve, thereby achieving rapid adjustment of the outlet NOx concentration.

[0010] (2) This invention constructs a prediction-feedback composite control framework. The main feedback control loop uses the deviation between the NOx concentration at the filter outlet and the set value as the adjustment basis to achieve closed-loop tracking and disturbance suppression of the NOx concentration at the outlet. The prediction compensation control loop uses the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet as the compensation basis, and performs equivalent correction on the phase lag caused by the time delay and higher-order inertia of the object under the action of the order reduction compensation link. Thus, under complex disturbance conditions such as rapid load changes, inlet NOx fluctuations and measurement delays, the overshoot, oscillation and slow response problems caused by the time delay of the control loop are significantly reduced.

[0011] (3) The present invention further introduces an adaptive tuning mechanism for controller parameters and a consistency verification mechanism for model parameters: At the control level, the controller parameters are adaptively calculated and tuned online according to the SCR dynamic characteristic model parameters under the current working conditions, so that the controller can automatically adjust the control strength and integral action as the working conditions such as smoke temperature and smoke volume drift, improve the robustness across working conditions and avoid the performance degradation caused by fixed parameter tuning; At the modeling level, the model parameter set identified under different working conditions is subjected to series consistency verification and joint correction, and a preset threshold is used as the entry criterion to ensure the effectiveness and reusability of the model parameter library. Attached Figure Description

[0012] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood in conjunction with the following description of the embodiments, in which: Figure 1 This is a flowchart of the control method of the present invention; Figure 2 This is an architecture diagram of the control method of the present invention; Figure 3 This is a flowchart of the model parameter identification and model parameter database construction process in this invention; Figure 4 This is a schematic diagram of the prediction-feedback composite control framework in this invention; Figure 5 This is a schematic diagram comparing the control performance curves of Embodiment 2 in this invention. Detailed Implementation

[0013] This invention aims to provide a method and apparatus for SCR denitrification time-delay compensation prediction-feedback composite control based on characteristic flue gas parameters. To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. The described embodiments are some, but not all, embodiments of this invention, and are exemplary, intended to explain the invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0014] Example 1: Control Method See Figure 1 , Figure 2 The present invention provides a SCR denitrification time-delay compensation prediction-feedback composite control method based on characteristic flue gas parameters. This method is used in SCR denitrification systems of thermal power generating units to generate ammonia injection valve opening control commands based on the outlet NOx concentration setpoint under varying operating conditions, thereby achieving precise adjustment and dynamic optimization of outlet NOx emissions. The method mainly includes the following steps in its implementation: SS1. Model Parameter Database Construction: Historical operating data of the SCR denitrification system are collected, and the corresponding SCR dynamic characteristic model parameters are identified under different characteristic flue gas parameter conditions. The mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters is established and a model parameter database is formed. This model parameter database is used to provide searchable parameter benchmarks for online operating condition matching, parameter interpolation / extrapolation and controller tuning.

[0015] Historical operating data should include time-series data on flue gas temperature, flue gas flow rate, ammonia injection valve opening, ammonia injection rate, and outlet NOx concentration of the SCR denitrification system. Before parameter identification, the data should undergo time alignment, outlier removal, and unit standardization. Time alignment should include unifying the sampling timescales of valve opening, ammonia injection rate, and NOx measurement signals, and applying necessary delay corrections to reduce the impact of differences in the data acquisition chain on the identification results. Characteristic flue gas parameters should include at least flue gas flow rate and flue gas temperature, used to characterize the current operating conditions of the SCR denitrification system and serve as index variables for establishing the mapping relationship between parameters and SCR dynamic characteristic model parameters. Flue gas flow rate reflects the spatial velocity and residence time characteristics of the catalyst bed, while flue gas temperature determines the catalytic reaction activity and denitrification efficiency level.

[0016] In this embodiment of the invention, the dynamic characteristic model of SCR under different characteristic flue gas parameters is described. G ( sThis includes: a sub-model of the dynamic characteristics of outlet NOx concentration with ammonia injection rate as input and outlet NOx concentration as output, and containing a time lag element. G 1( s Its transfer function satisfies ; and a sub-model of the dynamic characteristics of ammonia flow rate with ammonia injection valve opening as input and ammonia injection quantity as output. G 2( s Its transfer function satisfies The two are connected in series to form ,in s For the Laplace operator; K 1 represents the equivalent gain coefficient characterizing the effect of changes in ammonia injection rate on changes in outlet NOx concentration. K 2 represents the equivalent gain coefficient characterizing the effect of changes in the ammonia injection valve opening on changes in the ammonia injection quantity; T 1. T 2 represents the second-order inertial parameter used to characterize the dynamic response of the outlet NOx concentration. T 3 represents the first-order inertial time constant used to characterize the dynamic response of ammonia flow rate; τ The time delay is used to characterize the equivalent delay in the NOx concentration at the outlet after the change in the amount of ammonia injected, through injection mixing, transport and catalytic reaction. For example, it is the superposition effect of multiple physical processes such as the transport delay of ammonia from the injection point to the catalyst inlet, the diffusion delay inside the catalyst bed and the sensing delay of the NOx concentration measurement point at the outlet.

[0017] In this embodiment of the invention, parameter identification and model parameter database construction are performed for SCR dynamic characteristic models under different characteristic flue gas parameter conditions, such as... Figure 3 As shown, it includes the following sub-steps: S101. Operating condition sample division: Based on flue gas temperature and flue gas flow rate, historical operating data is divided into multiple operating condition sample sets, and candidate identification data segments are extracted in each operating condition sample set where the opening degree of the ammonia injection valve or the amount of ammonia injection changes and the outlet NOx concentration shows an identifiable dynamic response. Preferably, in step S101, the operating condition sample set is constructed by partitioning and discretizing the flue gas temperature and flow rate. The partitioning method involves dividing the flue gas temperature and flow rate into several preset intervals and forming an operating condition index with interval numbers. Furthermore, the candidate identification data segment satisfies at least one admission condition: the change amplitude of the ammonia injection valve opening or ammonia injection quantity is greater than a preset excitation threshold, and after the change occurs, the outlet NOx concentration exhibits a monotonically changing or significantly deviating from the steady state dynamic response within a preset time window, to ensure that subsequent time delay estimation and parameter identification are observable and repeatable. Further, the preset time window can be adaptively adjusted according to changes in flue gas flow rate to avoid response truncation due to an excessively short window under high flow rate conditions.

[0018] S102. Initial Time Delay Estimation: Analyze the time-series relationship between the change in ammonia injection rate and the response of NOx concentration at the outlet within each candidate identification data segment to obtain the time delay. τ The preliminary estimate is determined by the time delay corresponding to the maximum value of the cross-correlation function, or by the moment when the outlet NOx concentration first shows a significant change after a step change in the ammonia injection rate. Among them, the obvious change criterion is that the absolute value of the first difference of the outlet NOx concentration exceeds the noise level threshold and the duration exceeds two sampling periods; in addition, in the cross-correlation function method, the normalized cross-correlation coefficient is used as a similarity measure, and the delay time corresponding to the global maximum value is found within the time delay search range; in the step response method, the response start point is identified by performing sliding window variance detection on the outlet NOx concentration signal, and the time difference between this moment and the step moment of ammonia injection is calculated as the initial value of the time delay.

[0019] S103. Model parameter identification: Given the time delay... τ Based on this, the least squares method or recursive least squares method is used to identify the model parameters, with the ammonia injection valve opening as the input and the ammonia injection quantity as the output. G 2( s ) parameters K 2 and T 3. Identification using ammonia injection rate as input and outlet NOx concentration as output. G 1( s ) parameters K 1. T 1. T 2, and during the identification process, parameters T 1. T 2. T 3. Applying positive and τ For non-negative physical constraints; S104. Series Consistency Check: Perform joint correction on the identified set of model parameters to ensure... G ( s Under the same operating conditions, the dynamic response fitting error of the outlet NOx concentration is minimized, and the preset threshold is used as the warehousing criterion. For the identification results that do not meet the warehousing criterion, steps S102~S103 are re-executed. Preferably, in step S104, the fitting error is an error index calculated based on the difference between the measured outlet NOx concentration and the model output, including the root mean square error (RMSE) and / or the normalized mean square error (NMSE). The inclusion criterion is that the error index is not greater than a preset threshold, and the joint correction uses the error index as the objective function to perform constrained optimization calculations on the model parameter set. When the error index after joint correction is still greater than the preset threshold, steps S102 to S103 are re-executed to update the initial time delay value and model parameters until the inclusion criterion is met or the preset retry limit is reached.

[0020] S105. Parameter Record Generation and Database Writing: When the fitting error meets the database entry criteria, a mapping relationship is established between the flue gas temperature and flue gas flow rate index variables corresponding to each operating condition sample set and the parameter set of the jointly corrected SCR dynamic characteristic model, and written into the model parameter database. The mapping relationship is stored in the form of a multidimensional lookup table or parameterized function, supporting subsequent fast retrieval and interpolation calculation. The database record format includes operating condition index (flue gas temperature interval number, flue gas flow rate interval number), model parameters (K1, K2, T1, T2, T3, τ, etc.), and identified quality indicators (RMSE, NMSE, R). 2 The system includes fields such as data segment timestamp and data segment sample quantity, and assigns a unique identifier to each record for traceability and updating.

[0021] SS2. Current operating condition model determined: The characteristic flue gas parameters of the current operating condition are collected in real time. Based on the model parameter database and the mapping relationship, the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters are obtained, and the SCR dynamic characteristic model under the current operating condition is generated and used as the subsequently controlled object.

[0022] In this embodiment of the invention, obtaining the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters based on the mapping relationship includes: calculating the distance metric between the current characteristic flue gas parameters and the characteristic flue gas parameters of each operating condition interval in the model parameter database, and selecting the parameter group with the smallest distance as the current model parameters. The distance metric can be the Euclidean distance or weighted distance after dimension normalization of flue gas temperature and flue gas volume to avoid a single dimension dominating the matching result; if the current characteristic flue gas parameters fall within the overlapping range of adjacent operating condition intervals, weighted interpolation and fusion of at least two sets of model parameters are performed based on the distance weight to generate model parameters; and if the current characteristic flue gas parameters exceed the coverage range of the database, at least two candidate parameter groups with the smallest distance are selected, and the corresponding model parameters are obtained by linear extrapolation calculation. During linear extrapolation calculation, physical rationality constraints need to be applied to the extrapolation results, limiting the parameter change gradient to no more than a preset multiple (e.g., 1.5 times) of the parameter change gradient of adjacent operating condition intervals within the database, and outputting an operating condition extrapolation warning flag.

[0023] SS3. Predicted Output Generation: Based on the SCR dynamic characteristic model under the current operating conditions, a prediction model is constructed to predict the outlet NOx response under ammonia injection control to obtain the predicted outlet NOx concentration. The real-time acquired outlet NOx concentration signal is then low-pass filtered to obtain the filtered outlet NOx concentration, providing necessary state information and feedforward predictions for the prediction-feedback composite control framework.

[0024] In this embodiment of the invention, the prediction model adopts the SCR dynamic characteristic model under the current operating conditions. Consistent transfer function form ,in G 0( s () represents the equivalent dynamic transmission link without time delay in the SCR dynamic characteristic model. G m ( s This refers to the predictive dynamic process without time delay. τ m To estimate the time delay, and considering the settings of each parameter in the estimation model and the controlled object... G ( s Maintaining consistency in dynamic characteristics; the forward projection of future outlet NOx concentrations is achieved by explicitly introducing a time-delay compensation mechanism in the prediction calculation, including: within each control cycle, the ammonia injection valve opening control quantity is used as the input to the prediction model, first through... G m ( s The intermediate estimated output without time delay is calculated. P m ( s Then, based on the estimated time delay, the process is adjusted accordingly. P m ( s An equivalent time delay is applied to obtain the estimated outlet NOx concentration, which makes the estimated outlet NOx concentration lead the filtered outlet NOx concentration in time, providing a time delay compensation prediction for the prediction compensation control loop and reducing the regulation lag caused by the time delay.

[0025] In addition, a first-order low-pass filter is used for low-pass filtering. The real-time acquired outlet NOx concentration signal is low-pass filtered, with a filtering time constant of... τ f The system adaptively adjusts the NOx concentration at the outlet based on the measured noise level and control cycle, ensuring that the NOx concentration at the filtered outlet maintains observability of operating disturbances while suppressing high-frequency noise. The filtered signal is then used as one of the common inputs to both the main feedback control loop and the predictive compensation control loop.

[0026] SS4. Construction of a Prediction-Feedback Composite Control Framework: Construct a prediction-feedback composite control framework, such as Figure 4 As shown, it includes a main feedback control loop and a predictive compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the set value of the NOx concentration at the outlet. The predictive compensation control loop generates the predictive compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet through order reduction compensation processing.

[0027] In this embodiment of the invention, the main feedback control loop includes a first controller. G c1 ( s ) and the controlled object G ( s ), G c1 ( s The input to the filter outlet NOx concentration is the deviation signal between the filtered outlet NOx concentration and the setpoint, and the output is the main feedback control quantity. When the output of the main feedback control loop becomes saturated, anti-saturation processing is performed on its internal integral state to avoid secondary overshoot of the outlet NOx caused by integral accumulation. The prediction compensation control loop includes a second controller. G c2 ( s ), downgrade compensation stage F 2( s ) and prediction models, G c2 ( s The input to the filter is the deviation signal between the estimated outlet NOx concentration and the filtered outlet NOx concentration. The output is processed by a step-down compensation circuit. F 2( s After processing, an estimated compensation control quantity is generated; F 2( s Used for controlling the object G ( s The higher-order inertia and phase lag caused by pure time delay are equivalently compensated, so that the predictive compensation control loop can enhance the dynamic response and phase margin of the composite control framework under varying operating conditions.

[0028] Degradation compensation stage F 2( s By constructing a dynamic characteristic model of SCR under current operating conditions G ( s A matching dynamic correction network is implemented, whose transfer function satisfies ,in a This is the order reduction compensation intensity coefficient, used to characterize the degree of reconstruction of the equivalent inertial order and response velocity. ξThe damping ratio is used to constrain the oscillation tendency of the compensation circuit and improve its robust stability; and the order of the compensation circuit is reduced. F 2( s The low-frequency gain remains unchanged and is within the time delay period. τ Provides phase correction within the corresponding frequency band, enabling G c2 ( s The output of the predicted compensation control quantity can reduce the adverse effects of time delay and inertia on the dynamic performance of the composite control framework.

[0029] G c1 ( s )and G c2 ( s All are PI controllers, and their transfer functions all satisfy... ,in K p For proportional gain, T i The integral time constant is denoted as ; and G c1 ( s )and G c2 ( s Configure independent sets of proportional gain and integral time constant parameters respectively, so that G c1 ( s The system implements closed-loop regulation for tracking and suppressing the setpoint for NOx concentration at the outlet. G c2 ( s The deviation between the estimated outlet NOx concentration and the filtered outlet NOx concentration is compensated and adjusted to improve the adaptability of the composite control framework to the effects of time delay and inertia.

[0030] SS5. Controller parameter adaptive tuning: Based on the SCR dynamic characteristic model parameters under current operating conditions, adaptive calculation and tuning are performed on the controller parameters in the main feedback control loop and the predictive compensation control loop. This ensures that the main feedback control loop meets stability requirements and achieves closed-loop tracking and disturbance suppression of the outlet NOx concentration. The predictive compensation control loop achieves time delay compensation and dynamic response optimization, reducing the impact of time delay and operating condition changes on the closed-loop dynamic performance. The adaptive calculation and tuning of controller parameters includes at least the following sub-steps: Based on the SCR dynamic characteristic model gain and main inertia time constant obtained in step SS2, the proportional and integral parameters of the main feedback control loop are calculated. Based on the time delay parameters, the compensation intensity and action time scale of the estimated compensation control loop are calculated. After tuning, the closed-loop stability margin is judged. When the stability margin does not meet the preset conditions, the bandwidth of the main feedback control loop is reduced and the output change rate of the estimated compensation control loop is limited. The stability margin can be characterized by the phase margin and / or gain margin to ensure that the closed loop remains stable when there are operating disturbances and model mismatches. Furthermore, during the adaptive tuning process, online consistency verification is performed. The estimated outlet NOx concentration obtained in step SS3 and the filtered outlet NOx concentration are statistically analyzed within the same time window. If the mean or variance of the error exceeds the threshold, it is determined that there is a mismatch in the model parameters under the current operating condition, and the model parameter reselection or reinterpolation process is triggered. At the same time, a smooth update constraint is applied to the controller parameters to ensure that the change in the control parameters in adjacent control cycles does not exceed the preset upper limit.

[0031] SS6. Synthesis of Composite Control Quantities: The main feedback control quantity and the estimated compensation control quantity are superimposed to form the ammonia injection composite control quantity, and then converted into an ammonia injection valve opening control command based on the correspondence. The correspondence includes at least the static mapping relationship between the ammonia injection composite control quantity and the ammonia injection valve opening. Temperature constraints and flow constraints are superimposed during the conversion process to limit the upper limit of the ammonia injection valve opening command when the SCR inlet flue gas temperature deviates from the effective temperature range of the catalytic reaction or when the flue gas flow fluctuates drastically, thereby reducing the risk of ammonia escape and ammonium salt deposition.

[0032] SS7. Control Command Output and Loop Update: The ammonia injection valve opening control command is output to the ammonia injection valve actuator of the SCR denitrification system, and steps SS2 to SS6 are executed cyclically according to a preset control cycle to continuously update the controlled object and output the ammonia injection valve opening control command under varying operating conditions. The cyclic update includes the validity judgment of key measurement signals. When the outlet NOx concentration signal, characteristic flue gas parameter signal, or ammonia injection valve feedback signal meets preset anomaly criteria, the model update in step SS2 is frozen, and steps SS3 to SS6 are continued using the most recently valid SCR dynamic characteristic model and controller parameters. Simultaneously, a fault flag is output to trigger the operation protection strategy. The anomaly criteria include at least one of the following: signal out of bounds, mutation rate exceeding limits, prolonged freeze, or communication interruption, to improve the fault tolerance of online operation.

[0033] It should be noted that this embodiment achieves adaptive updating of the dynamic characteristic model of the SCR denitrification system under operating conditions by constructing a model parameter database based on characteristic flue gas parameters, thus solving the problem of model mismatch in a wide operating range of traditional fixed parameter control methods. By constructing a predictive-feedback composite control framework, the steady-state accuracy advantage of the main feedback control loop and the dynamic advance advantage of the predictive compensation control loop are organically combined, effectively compensating for the large time delay and high-order inertial characteristics of the controlled object, significantly improving the response speed and robustness of outlet NOx concentration control, and providing an effective technical means for thermal power generating units to achieve ultra-low emissions under deep peak shaving and rapid load change conditions.

[0034] It should also be noted that this invention improves the convergence and reliability of parameter identification by decomposing the SCR dynamic characteristic model into a series structure of an ammonia flow quantum model and an outlet NOx concentration sub-model, and by combining independent identification and joint correction of the sub-models. The introduction of the order reduction compensation stage achieves equivalent order reduction and phase lead compensation for the high-order system, expanding the bandwidth of the control system while ensuring closed-loop stability. The adaptive tuning and online consistency verification mechanism of the controller parameters ensures smooth switching and continuous optimization of the composite control framework during changes in operating conditions, avoiding control performance degradation caused by parameter mutations. The dynamic superposition strategy of temperature constraints and flow constraints effectively suppresses the risk of ammonia escape while ensuring denitrification efficiency. In the composite control framework, this invention uses the predicted output to form lead information that can be used for compensation, and achieves a balance between phase margin and response speed across operating conditions through order reduction compensation and adaptive tuning, providing a systematic means to suppress complex factors such as object drift, load mutations, and measurement delays during long-term operation.

[0035] Example 2: Control Device Based on Embodiment 1 above, Embodiment 2 provides a SCR denitrification time delay compensation prediction-feedback composite control device based on characteristic flue gas parameters, corresponding to the above control method, including: The model parameter database module is used to store the SCR dynamic characteristic model parameters identified from historical operating data under different characteristic flue gas parameter conditions, as well as the mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters. The current operating condition model generation module is used to receive characteristic flue gas parameters in real time, obtain the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters based on the model parameter database and in combination with the mapping relationship, and generate the SCR dynamic characteristic model under the current operating condition as the controlled object. The predicted output generation module is used to build a prediction model based on the SCR dynamic characteristic model under the current operating conditions, to predict the outlet NOx concentration by predicting the outlet NOx response under ammonia injection control, and to perform low-pass filtering on the real-time outlet NOx concentration to obtain the filtered outlet NOx concentration. The composite control framework module includes a main feedback control loop and a predictive compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the set value of the NOx concentration at the outlet. The predictive compensation control loop generates the predictive compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet, combined with the order reduction compensation processing. The parameter adaptive tuning module is used to adaptively calculate and tune the controller parameters in the main feedback control loop and the predictive compensation control loop based on the SCR dynamic characteristic model parameters under the current operating conditions. The composite control quantity synthesis module is used to superimpose the main feedback control quantity and the estimated compensation control quantity to form the ammonia injection composite control quantity, and convert it into the ammonia injection valve opening control command. The cyclic update module is used to drive each module to execute cyclically according to a preset control cycle, so as to continuously update the controlled object and continuously output the ammonia injection valve opening control command under changing operating conditions.

[0036] Example 3: Application Case Based on Examples 1 and 2 above, Example 3 uses a 300MW coal-fired power generating unit's SCR denitrification system as an application example to detail the implementation process, parameter tuning method, and control effect verification of the control method of the present invention in a practical engineering project. This unit is equipped with a honeycomb catalyst SCR denitrification system, with the catalyst arranged in a 2+1 layer structure, and a designed flue gas treatment capacity of 1,100,000 Nm³. 3 / h, the design value for inlet NOx concentration is 450 mg / Nm³. 3 The target for NOx concentration control at exports is 50 mg / Nm³. 3 the following.

[0037] Approximately 15 million operational data points were collected from January to June 2024, with a sampling frequency of 1 second. After preprocessing including time alignment, outlier removal (3σ criterion), and missing value imputation, 25 operating condition grids were constructed based on flue gas temperature and flow rate. A sliding window method (window length 600 seconds, step size 60 seconds) was used to screen data segments that met the admission criteria, resulting in 1247 valid data segments and a database coverage rate of 92%.

[0038] Operating conditions [360~380℃] × [880000~990000 Nm] 3For example, the data segment where the ammonia injection valve opening jumps from 45% to 52% is selected, and parameter identification is carried out according to the SS2 method: (1) Time delay identification: the time delay time is determined to be τ=111 seconds by using the cross-correlation function method; (2) Ammonia gas flow quantum model identification: the least squares method is used to obtain K2=7.76 kg / h / %, T3=58.51 seconds, R 2 =0.94, forming (3) Sub-model identification of NOx concentration at the outlet: K1 = -1.547 mg / Nm was obtained by recursive least squares method. 3 / (kg / h), T1=86.02 seconds, T2=42.44 seconds, R 2 =0.91, forming (4) Series consistency check: RMSE = 1.8 mg / Nm 3 The NMSE value is 0.078, which meets the warehousing criterion. The final SCR dynamic characteristic model under this operating condition is... .

[0039] After identifying, verifying and determining whether to include each of the 25 operating conditions according to the above process, 23 sets of valid model parameters were finally obtained and mapped to the corresponding flue gas temperature and flue gas flow operating condition indexes. These parameters were then written into the model parameter database for online operating condition matching, interpolation / extrapolation and adaptive updating of controller parameters.

[0040] The operating conditions obtained from the aforementioned identification are [360~380℃] × [880000~990000 Nm]. 3 The SCR dynamic characteristic model under / h] calculates the controller parameters according to the adaptive tuning method in step SS5 of Example 1.

[0041] Low-pass filter design. Low-pass filter transfer function selection. The noise standard deviation σ, based on the NOx concentration at the outlet, is 0.8 mg / Nm³. 3 With the dominant time constant T2 = 42.44 seconds, τ f =0.5×T2=21.2 seconds, discretization filter coefficient α=Δt / (τ f +Δt)=1 / (21.2+1)=0.045.

[0042] The controlled object is a third-order system (n=3), which is equivalent to a first-order system using a second-order reduction compensation element F2(s). Following the formula in Example 1, with a=3 and ξ=0.5, the compensator time constant τ′ is calculated: Second The equivalent first-order time constant is: Second; The transfer function of the order reduction compensation stage is: Main feedback controller G c1 Parameter tuning. The equivalent first-order model G′(s)=K′ / (1+T′s) is used for tuning the main feedback controller, where K′=K1·K2=-12 mg / Nm. 3 T′ = 99.32 seconds. Using the SIMC tuning rules, the desired closed-loop time constant λ1 = 2 × τ = 2 × 111 = 222 seconds is obtained from the calculation. K p1 Taking the absolute value of 0.037, the negative sign is reflected in the control loop connection method, that is, the set value minus the measured value constitutes a positive deviation. Simultaneously, calculation... Seconds, therefore, the main feedback controller is constructed as Seconds. Considering practical debugging experience, to improve system response speed, the proportional gain was fine-tuned, and K was ultimately chosen. p1 =0.08, integration time constant T i1 =80 seconds, get .

[0043] Predictive compensation controller G c2 Parameter tuning. The compensation gain proportional coefficient β is set to 0.5, and the estimated proportional gain K of the compensation controller is determined. p2 =β·K p1 =0.5 × 0.08 = 0.04. Integration time constant T i2 =T2 / 2=42.44 / 2=21.22 seconds. Closed-loop verification shows that under this parameter combination, the step response overshoot is 18%, the settling time is 320 seconds, the phase margin PM=42°, and the gain margin GM=8.5dB, meeting the stability margin requirements. To further optimize dynamic performance, the integral time constant is adjusted, and K is finally determined. p2 =0.08, T i2 =280 seconds, therefore With this parameter combination, the step response overshoot is reduced to 12%, the settling time is shortened to 280 seconds, the phase margin PM=48°, and the amplitude margin GM=10.2dB, resulting in a significant improvement in control performance.

[0044] The composite control framework is deployed on the industrial PLC and the power plant DCS, exchanging signals via a standard communication interface: uplink acquires measurements such as flue gas temperature, flue gas flow rate, and outlet NOx, while downlink outputs ammonia injection valve opening commands; and a control cycle and signal validity discrimination mechanism are set. When characteristic flue gas parameters change across operating conditions, the control system switches / interpolates and updates model parameters according to the model parameter database and synchronously adjusts the controller parameters, realizing proactive adjustment and closed-loop correction of the ammonia injection valve under varying operating conditions.

[0045] To verify the control performance, two types of tests were set up under the same operating conditions: one was to apply a step change to the outlet NOx concentration setpoint at t=100s; the other was to apply an equivalent load / inlet disturbance from t=2550 to 2950s to simulate the outlet NOx shift caused by a sudden change in operating conditions. Three control strategies were compared: the predictive-feedback composite control strategy of this invention, the conventional Smith compensation control strategy, and the conventional cascade control strategy. All three operated under the same sampling period and the same valve amplitude / rate of change constraints.

[0046] Table 1 Comparison of key indicators during the setpoint step phase (corresponding to) Figure 5 (Step jump around t=100 s) As shown in Table 1 and Figure 5 As shown, for a step jump around t=100s, the present invention has a smaller overshoot after the step jump and enters the stable region faster; compared with conventional Smith compensation control, the present invention effectively suppresses the peak impact caused by large time delay; compared with conventional cascade control, the present invention significantly shortens Ts and reduces IAE while ensuring a smaller overshoot, demonstrating better overall dynamic quality.

[0047] Table 2 Comparison of key indicators in the disturbance suppression phase (corresponding to) Figure 5 (Disturbance segment t=2550~2950 s)

[0048] As shown in Table 2 and Figure 5 As shown, for Figure 5 During the disturbance segment from t=2550 to 2950s, the maximum deviation of the present invention is smaller, the recovery is faster, and the overall error IAE is significantly reduced. This indicates that the present invention can reduce hysteresis accumulation and overcorrection in both the two key transients of disturbance injection and release through the synergistic effect of predicting lead amount, feedback closed-loop correction, and order reduction compensation phase correction, thereby achieving stronger disturbance suppression capability.

[0049] The objectives of this invention have been fully and effectively achieved through the above embodiments. Those skilled in the art will understand that this invention includes, but is not limited to, the contents described in the accompanying drawings and the specific embodiments described above. Although the invention has been described with reference to what is currently considered the most practical and preferred embodiments, it should be understood that the invention is not limited to the disclosed embodiments, and any modifications that do not depart from the functional and structural principles of the invention will be included within the scope of the claims.

Claims

1. A composite control method for SCR denitrification time delay compensation prediction-feedback based on characteristic flue gas parameters, characterized in that, It should include at least the following steps: SS1. Collect historical operating data of the SCR denitrification system, identify the corresponding SCR dynamic characteristic model parameters under different characteristic flue gas parameter conditions, and form a model parameter database; SS2. Real-time acquisition of characteristic flue gas parameters under the current operating conditions, acquisition of model parameters corresponding to the current characteristic flue gas parameters based on the model parameter database, and generation of SCR dynamic characteristic model; SS3. Based on the SCR dynamic characteristic model under the current operating conditions, a prediction model is constructed to predict the NOx response at the outlet to obtain the predicted NOx concentration at the outlet. The real-time collected NOx concentration signal at the outlet is then low-pass filtered to obtain the filtered NOx concentration at the outlet. SS4. Construct a prediction-feedback composite control framework, which includes a main feedback control loop and a prediction compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the setpoint of the NOx concentration at the outlet. The prediction compensation control loop generates the prediction compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet through order reduction compensation processing. SS5. Tune the main feedback control loop and the predictive compensation control loop to ensure that the main feedback control loop meets the stability requirements and that the predictive compensation control loop achieves time delay compensation and dynamic response optimization. SS6. The main feedback control quantity and the estimated compensation control quantity are superimposed to form the ammonia injection composite control quantity, and converted into an ammonia injection valve opening control command based on the corresponding relationship; SS7. Output the ammonia injection valve opening control command to the ammonia injection valve actuator of the SCR denitrification system, and execute steps SS2~SS6 in a cycle according to the preset control period.

2. The method according to claim 1, characterized in that, In step SS1, the historical operating data includes at least time series data of flue gas temperature, flue gas flow rate, ammonia injection valve opening, ammonia injection quantity, and outlet NOx concentration of the SCR denitrification system. Before parameter identification, the data undergoes time alignment, outlier removal, and unit consistency processing. The characteristic flue gas parameters include at least flue gas flow rate and flue gas temperature, which are used to characterize the current operating conditions of the SCR denitrification system and serve as operating condition index variables for establishing the mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters.

3. The method according to claim 2, characterized in that, In step SS1, the dynamic characteristic model of SCR under different characteristic flue gas parameters. G ( s This includes a sub-model of the dynamic characteristics of outlet NOx concentration, which takes ammonia injection rate as input, outlet NOx concentration as output, and contains a time lag element. G 1( s Its transfer function satisfies And a sub-model of the dynamic characteristics of ammonia flow rate with ammonia injection valve opening as input and ammonia injection quantity as output. G 2( s Its transfer function satisfies The two sub-models are connected in series to form G ( s And satisfy ,in s For the Laplace operator; K 1. K 2 represents the equivalent gain coefficient; T 1. T 2. T 3 represents the inertial time constant; τ This refers to the time lag.

4. The method according to claim 3, characterized in that, Step SS1, which involves parameter identification and model parameter database construction for the SCR dynamic characteristic model, includes: S101. Operating condition sample division: Based on flue gas temperature and flue gas flow rate, historical operating data is divided into multiple operating condition sample sets, and candidate identification data segments are extracted in each operating condition sample set where the opening degree of the ammonia injection valve or the amount of ammonia injection changes and the outlet NOx concentration shows an identifiable dynamic response. S102. Initial Time Delay Estimation: Analyze the time-series relationship between the change in ammonia injection rate and the response of NOx concentration at the outlet within each candidate identification data segment to obtain the time delay. τ The preliminary estimate is determined by the time delay corresponding to the maximum value of the cross-correlation function, or by the moment when the outlet NOx concentration first shows a significant change after a step change in the ammonia injection rate. S103. Model parameter identification: Given the time delay... τ Based on this, the least squares method is used to identify the model parameters, with the ammonia injection valve opening as the input and the ammonia injection quantity as the output. G 2( s ) parameters K 2 and T 3. Identification using ammonia injection rate as input and outlet NOx concentration as output. G 1( s ) parameters K 1. T 1. T 2, and during the identification process, parameters T 1. T 2. T 3. Applying positive and τ For non-negative physical constraints; S104. Series Consistency Check: Perform joint correction on the identified set of model parameters to ensure... G ( s Under the same operating conditions, the dynamic response fitting error of the outlet NOx concentration is minimized, and the preset threshold is used as the warehousing criterion. For the identification results that do not meet the warehousing criterion, steps S102~S103 are re-executed. S105. Parameter Record Generation and Database Writing: When the fitting error meets the database entry criteria, establish a mapping relationship between the flue gas temperature and flue gas flow rate index variables corresponding to each working condition sample set and the jointly corrected SCR dynamic characteristic model parameter set, and write them into the model parameter database.

5. The method according to claim 1, characterized in that, In step SS2, obtaining the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters includes: Calculate the distance metric between the current characteristic flue gas parameters and the characteristic flue gas parameters of each operating condition interval in the model parameter database, and select the parameter group with the smallest distance as the current model parameters; if the current characteristic flue gas parameters fall within the overlapping range of adjacent operating condition intervals, perform weighted interpolation fusion on at least two groups of model parameters based on distance weight to generate model parameters; if the current characteristic flue gas parameters exceed the coverage range of the database, select at least two candidate parameter groups with the smallest distance, and obtain the corresponding model parameters through linear extrapolation.

6. The method according to claim 3, characterized in that, In step SS3, the prediction model Adopting the SCR dynamic characteristic model under current operating conditions A transfer function with consistent structure, wherein G 0( s )for G ( s The equivalent dynamic transmission link in the () does not contain time delay. G m ( s This refers to the predictive dynamic process without time delay. τ m To estimate the time delay, and the settings of each parameter in the estimation model are... G ( s Maintain consistency in dynamic characteristics; Forecast The calculation employs an explicit time-delay compensation mechanism to forward extrapolate the outlet NOx concentration at future moments. This includes: within each control cycle, using the ammonia injection valve opening control value as the input to the prediction model, first through... G m ( s The intermediate estimated output without time delay is calculated. P m ( s Then, based on the estimated time delay links... right P m ( s The estimated export NOx concentration is obtained by applying an equivalent time delay.

7. The method according to claim 6, characterized in that, In step SS3, the low-pass filtering process uses a first-order low-pass filter. The real-time acquired outlet NOx concentration signal is low-pass filtered, with a filtering time constant of... τ f The control cycle is adaptively adjusted based on the measured NOx level at the outlet, and the filtered signal is used as one of the common inputs to the main feedback and predictive compensation control loops.

8. The method according to claim 7, characterized in that, In step SS4, the main feedback control loop includes a first controller. G c1 ( s ) and the controlled object G ( s ), G c1 ( s The input to the filter outlet NOx concentration is the deviation signal between the filtered outlet NOx concentration and the setpoint value. The output is the main feedback control quantity, and anti-saturation processing is performed on its internal integral state when the output saturation occurs in the main feedback control loop. The prediction compensation control loop includes a second controller. G c2 ( s ), downgrade compensation stage F 2( s ) and prediction models, G c2 ( s The input to the filter is the deviation signal between the estimated outlet NOx concentration and the filtered outlet NOx concentration. The output is processed by a step-down compensation circuit. F 2( s After processing, an estimated compensation control quantity is generated; F 2( s Used for controlling the object G ( s The higher-order inertia and phase lag caused by pure time delay are compensated for equivalently.

9. The method according to claim 8, characterized in that, In step SS4, the order reduction compensation stage F 2( s By constructing a dynamic characteristic model of SCR under current operating conditions G ( s A matching dynamic correction network is implemented, whose transfer function satisfies ,in a For order reduction compensation strength coefficient, ξ The damping ratio; and the order reduction compensation stage. F 2( s The low-frequency gain remains unchanged and is within the time delay period. τ It provides phase correction within the corresponding frequency band.

10. The method according to claim 8, characterized in that, In step SS4, G c1 ( s )and G c2 ( s All are PI controllers, and their transfer functions all satisfy... ,in K p For proportional gain, T i The integral time constant is denoted by ; and G c1 ( s )and G c2 ( s Configure independent sets of proportional gain and integral time constant parameters respectively, so that G c1 ( s The system implements closed-loop regulation for tracking and suppressing the NOx concentration setpoint at the outlet. G c2 ( s The deviation between the estimated outlet NOx concentration and the filtered outlet NOx concentration is compensated and adjusted.

11. The method according to claim 1, characterized in that, In step SS5, the adaptive calculation and tuning of controller parameters includes at least: Based on the model gain and main inertia time constant obtained in step SS2, calculate the proportional and integral parameters of the main feedback control loop. Based on the time delay parameters, calculate the compensation intensity and action time scale of the estimated compensation control loop. After tuning, determine the closed-loop stability margin. When the stability margin does not meet the preset conditions, reduce the bandwidth of the main feedback control loop and limit the output change rate of the estimated compensation control loop. Furthermore, during the adaptive tuning process, online consistency verification is performed. The estimated outlet NOx concentration obtained in step SS3 and the filtered outlet NOx concentration are statistically analyzed within the same time window. If the mean or variance of the error exceeds the threshold, it is determined that there is a mismatch in the model parameters under the current operating condition, and the model parameter reselection or reinterpolation process is triggered. At the same time, a smooth update constraint is applied to the controller parameters.

12. The method according to claim 1, characterized in that, In step SS6, the correspondence includes at least the static mapping relationship between the ammonia injection composite control quantity and the ammonia injection valve opening degree. In the conversion process, flue gas temperature constraints and flow constraints are superimposed to limit the upper limit of the ammonia injection valve opening degree command when the flue gas temperature at the SCR inlet deviates from the effective temperature range of the catalytic reaction or when the flue gas flow fluctuates drastically.

13. The method according to claim 1, characterized in that, In step SS7, the cyclic update includes the validity judgment of key measurement signals. When the outlet NOx concentration signal, characteristic flue gas parameter signal, or ammonia injection valve feedback signal meets the preset abnormality criteria, the model update in step SS2 is frozen and the most recently valid SCR dynamic characteristic model and controller parameters are used to continue executing steps SS3 to SS6. At the same time, the fault flag quantity is output to trigger the operation protection strategy of the unit control system.

14. A composite control device for SCR denitrification time delay compensation prediction-feedback based on characteristic flue gas parameters, characterized in that, Includes the following modules: The model parameter database module is used to store the SCR dynamic characteristic model parameters identified from historical operating data under different characteristic flue gas parameter conditions, as well as the mapping relationship between characteristic flue gas parameters and SCR dynamic characteristic model parameters. The current operating condition model generation module is used to receive characteristic flue gas parameters in real time, obtain the SCR dynamic characteristic model parameters corresponding to the current characteristic flue gas parameters based on the model parameter database and in combination with the mapping relationship, and generate the SCR dynamic characteristic model under the current operating condition as the controlled object. The predicted output generation module is used to build a prediction model based on the SCR dynamic characteristic model under the current operating conditions, to predict the outlet NOx concentration by predicting the outlet NOx response under ammonia injection control, and to perform low-pass filtering on the real-time outlet NOx concentration to obtain the filtered outlet NOx concentration. The composite control framework module includes a main feedback control loop and a predictive compensation control loop. The main feedback control loop generates the main feedback control quantity based on the deviation between the NOx concentration at the filter outlet and the set value of the NOx concentration at the outlet. The predictive compensation control loop generates the predictive compensation control quantity based on the deviation between the predicted NOx concentration at the outlet and the NOx concentration at the filter outlet, combined with order reduction compensation processing. The parameter adaptive tuning module is used to adaptively calculate and tune the controller parameters in the main feedback control loop and the predictive compensation control loop based on the SCR dynamic characteristic model parameters under the current operating conditions. The composite control quantity synthesis module is used to superimpose the main feedback control quantity and the estimated compensation control quantity to form the ammonia injection composite control quantity, and convert it into the ammonia injection valve opening control command. The cyclic update module is used to drive each module to execute cyclically according to a preset control cycle, so as to continuously update the controlled object and continuously output the ammonia injection valve opening control command under changing operating conditions.

Images