Full-working-condition denitration system control method based on generalized predictive control and intelligent feedforward

Abstract

The invention provides an all-working-condition denitration system control method based on generalized predictive control and intelligent feedforward. The method comprises the following steps that a neural network algorithm is adopted, a transfer function model is obtained according to a historical data set and input into a generalized predictive controller, and a main loop controller is obtained; the main loop controller is used for main loop control to obtain a denitration inlet concentration prediction instruction, an ammonia spraying amount prediction instruction and denitration outlet concentration; and according to the ammonia injection amount prediction instruction, the denitration inlet concentration prediction instruction combined with the inlet concentration compensation algorithm, the denitration outlet concentration combined with the deviation protection logic, and the current ammonia injection amount combined with the historical optimal ammonia injection amount, obtaining a feed-forward master controller for controlling a feed-forward loop and updating the current ammonia injection amount. The control method provided by the invention has the advantages of good stability, reliability, predictive capacity, strong adaptability and good control quality, and can realize the optimal balance of denitration efficiency and economy.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal power generation, and in particular to a full-operating-condition denitrification system control method based on generalized predictive control and intelligent feedforward. Background Art

[0002] With increasingly stringent environmental regulations, thermal power plants face higher requirements for controlling nitrogen oxide (NOx) emissions. Selective catalytic reduction (SCR) denitrification technology is widely used due to its high efficiency. However, traditional control methods rely on fixed parameters or simple feedback, making them difficult to adapt to complex operating conditions such as load fluctuations and fuel changes. This leads to fluctuating denitrification efficiency, increased ammonia slip, and higher operating costs. Existing methods are limited in response speed and economic efficiency, making it impossible to balance environmental compliance with economic operation. Summary of the Invention

[0003] In response to the problems existing in the prior art, the present invention provides an intelligent denitrification control method based on generalized predictive control and multiple feedforward control modes, aiming to improve the denitrification efficiency and operating economy of thermal power generating units under all operating conditions.

[0004] To achieve the above-mentioned object, the technical solution adopted by the present invention is as follows: a control method for a full-operation-condition denitrification system based on generalized predictive control and intelligent feedforward, comprising the following steps: obtaining a historical data set, a current input instruction, and a current ammonia injection amount based on a thermal power generating unit; Using a neural network algorithm, a transfer function model is obtained based on the historical data set; The transfer function model is input into the generalized predictive controller to obtain the main loop controller; the optimization objective function of the main loop controller is: ; Where, To optimize the objective function, is the set value of denitrification outlet concentration, For time, is the denitrification outlet concentration, is the weight coefficient, is the change in the control variable; Based on the main loop controller, a prediction instruction is obtained according to the current input instruction; the prediction instruction at least includes a denitration inlet concentration prediction instruction and an ammonia injection amount prediction instruction; controlling the main circuit of the thermal power generating set according to the prediction instruction to obtain at least current output data, wherein the current output data at least includes a denitration outlet concentration; Based on the denitration inlet concentration prediction instruction, a first feedforward control is obtained according to an inlet concentration compensation algorithm; Based on the denitration outlet concentration, obtaining a second feedforward control according to a deviation protection logic; Obtaining a historical optimal ammonia injection amount according to the historical data set; obtaining a third feedforward control according to the current ammonia injection amount and the historical optimal ammonia injection amount; Obtaining a feedforward master controller according to the ammonia injection amount prediction instruction, the first feedforward control, the second feedforward control, and the third feedforward control; The feedforward loop of the thermal power generator set is controlled according to the feedforward master controller and the current ammonia injection amount is updated.

[0005] In some embodiments, the historical data set includes a first historical data set and a second historical data set; The first historical data set is used to obtain the transfer function model using a neural network algorithm; The second historical data set is used to obtain the historical optimal ammonia injection amount.

[0006] In some embodiments, the first historical data set includes a set of historical input data and a set of historical output data; The historical input data includes historical load, historical coal feed rate, historical ammonia injection flow rate, historical primary air damper opening and historical secondary air damper opening; The historical output data includes historical denitrification inlet concentration and historical denitrification outlet concentration.

[0007] In some embodiments, the current input command includes a load command and a coal feed rate command.

[0008] In some embodiments, the formula of the inlet concentration compensation algorithm is: ; Where, is the feedforward channel compensation value based on the predicted instruction, It is the delay time from judging the concentration change trend of denitrification inlet to the change of denitrification outlet concentration. is a function related to both load instruction and coal supply instruction, For coal feeding instruction, For load instructions.

[0009] In some embodiments, based on the denitration outlet concentration, the step of obtaining the second feedforward control according to the deviation protection logic is: Obtain sensitivity coefficient, deviation protection threshold and outlet concentration setting value; Obtaining a concentration change rate according to the denitration outlet concentration; Determining whether a first condition is met based on the denitration outlet concentration, the deviation protection threshold, and the outlet concentration setting value; the first condition is that the denitration outlet concentration is greater than the sum of the deviation protection threshold and the outlet concentration setting value; Determining whether a second condition is met based on the concentration change rate, the sensitivity coefficient, and the deviation protection threshold; the second condition is that the concentration change rate is greater than the product of the sensitivity coefficient and the deviation protection threshold; When both the first condition and the second condition are met, triggering an emergency control strategy; When the first condition is not met or the second condition is not met, the emergency control strategy is not triggered.

[0010] In some embodiments, the correction formula of the third feedforward control is: ; Where, To correct the signal, is the economic feedforward gain coefficient, This is the best ammonia injection amount in history. is the current ammonia injection amount.

[0011] In some embodiments, the overall feedforward controller is a weighted superposition of the output signal of the first feedforward control, the output signal of the second feedforward control, and the output signal of the third feedforward control.

[0012] In some embodiments, the second historical data set includes a set of system operating parameters of the thermal power generating set under different operating conditions, and historical transient ammonia injection amounts under different operating conditions.

[0013] In some embodiments, the feedforward master controller is used to control an ammonia injection regulating valve.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention, centered on generalized predictive control (GPC), builds a multi-level coordinated control system, combining inlet concentration compensation that fully accounts for the complex operating conditions and dynamic characteristics of thermal power generators, outlet concentration deviation protection based on environmental safety requirements, and economic feedforward analysis based on historical ammonia injection data. By running historical data modeling and neural network algorithms for system identification, an accurate transfer function model is formed, ensuring the control system possesses excellent predictive capabilities and adaptability.

[0015] 2. The present invention ensures precise control of the ammonia injection regulating valve through weighted superposition of multiple control signals, ensures that the environmentally friendly denitrification control system can achieve good control quality under all working conditions, and achieves the optimal balance between denitrification efficiency and economy.

[0016] 3. The present invention is applicable to denitrification control under complex working conditions of thermal power units and has good stability, reliability and promotion and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the principle of the full-condition denitrification system control method based on generalized predictive control and intelligent feedforward of the present invention. DETAILED DESCRIPTION

[0018] To clearly illustrate the technical features of this solution, the following detailed description of the implementation methods of this application will be given in conjunction with the accompanying drawings and examples, so that the implementation process of how this application applies technical means to solve technical problems and achieve corresponding technical effects can be fully understood and implemented accordingly. The embodiments of this application and the various features therein can be combined with each other as long as they do not conflict with each other, and the resulting technical solutions are all within the scope of protection of this application.

[0019] In order to achieve the goal of ensuring the economic efficiency of power plants while meeting environmental friendliness, it is urgent to develop a denitrification control method that is adaptable to all operating conditions, has predictive capabilities and intelligent adjustment functions, so as to improve the environmental performance and operating economy of thermal power units.

[0020] See also Figure 1 , an embodiment of the present invention provides a full-operation-condition denitrification system control method based on generalized predictive control and intelligent feedforward, comprising the following steps: obtaining a first historical data set, a second historical data set, a current input instruction, and a current ammonia injection amount based on a thermal power generating unit; The first historical data set is a collection of historical data of the operation of the thermal power generating unit; in some embodiments, the first historical data set includes a collection of historical input data and a collection of historical output data; the historical input data includes but is not limited to historical load, historical coal feed rate, historical ammonia injection flow rate, historical primary air damper opening, and historical secondary air damper opening; the historical output data includes historical denitrification inlet concentration and historical denitrification outlet concentration; Preprocessing the first historical data set to obtain a standard data set; preprocessing is used to remove industrial noise and coarse value data to ensure the accuracy and reliability of the data; Firstly, industrial noise is removed by wavelet packet processing, and then coarse value data is processed by quartile method; The steps for removing industrial noise using wavelet packet processing are as follows: Using wavelet packet decomposition, the collected data, which includes multi-dimensional operating parameters such as historical load, historical coal feed rate, historical ammonia injection flow rate, historical primary air damper opening, historical secondary air damper opening, historical denitrification inlet concentration, and historical denitrification outlet concentration, is decomposed into different frequency bands. A threshold denoising algorithm is then used to filter out high-frequency noise components. The signal is then reconstructed using wavelet packets, retaining the valid information as an intermediate data set. Compared to traditional filtering techniques, wavelet packet processing can more accurately separate noise from useful signals, reduce data distortion, and ensure high reliability and accuracy of data during the model identification phase.

[0021] The steps for processing outlier data using the quartile method are as follows: calculating the first quartile (Q1) and third quartile (Q3) of the intermediate dataset, determining the interquartile range (IQR) (Q3-Q1), and setting 1.5 times the IQR as the anomaly detection threshold. Any data in the intermediate dataset that falls below Q1-1.5IQR or exceeds Q3+1.5IQR is considered an outlier and is removed. This method of processing outlier data using the quartile method does not require assumptions about data distribution and is suitable for the variable operating conditions of thermal power generator operating data. It effectively improves data quality and ensures the stability and generalization of subsequent transfer function models.

[0022] The standard data set is classified according to the input-output relationship to obtain a training data set; the training data set includes a standard input data set and a standard output data set, the standard input data set is a set of preprocessed historical input data, and the standard output data set is a set of preprocessed historical output data; A neural network algorithm is used to obtain a transfer function model based on a training set. This algorithm is then used to identify the transfer function model of the system (thermal power generator) and establish a transfer function model that can characterize the system's dynamic characteristics. The model identification process includes selecting an appropriate network structure, determining a learning rate, and training and validating the transfer function model to ensure that the transfer function model has good generalization capabilities under different operating conditions. Using a neural network algorithm, modeling is performed based on the transfer function of the training set, ensuring that the resulting transfer function model effectively reflects the dynamic behavior of the thermal power generator and provides a reliable mathematical basis for overall control design.

[0023] The transfer function model is input into the generalized predictive controller to obtain the main loop controller. The denitrification outlet concentration is set as the feedback value of the main loop controller, and the denitrification outlet concentration set value is the target value of the main loop controller. The main loop controller realizes the preliminary control of the main loop of the thermal power generator set through the key links such as the transfer function model, rolling optimization and feedback correction in the generalized predictive controller. The optimization objective function of the main loop controller is: ; Where, To optimize the objective function, is the set value of denitrification outlet concentration, For time, is the denitrification outlet concentration, is the weight coefficient, is the change of the controlled variable, and the controlled variable is the amount of ammonia injected.

[0024] Based on the main loop controller, a prediction instruction is obtained according to the current input instruction; the prediction instruction includes at least a denitrification inlet concentration prediction instruction and an ammonia injection amount prediction instruction, and the prediction instruction also includes a denitrification outlet concentration instruction; in some embodiments, the current input instruction includes a load instruction and a coal feeding amount instruction, and by adjusting the prediction time domain and control time domain parameters, the control performance is optimized to ensure the system response speed and stability; the thermal power generating set includes at least a instruction input unit, a denitrification outlet and an ammonia injection regulating valve, and the instruction input unit is used to input current input instructions such as the load instruction and the coal feeding amount instruction.

[0025] According to the prediction instruction, the main circuit of the thermal power generating unit is controlled to obtain at least current output data, the current output data at least including the denitration outlet concentration, and the denitration outlet concentration monitored at the denitration outlet is controlled by the denitration outlet concentration instruction; Based on the denitration inlet concentration and the denitration inlet concentration prediction instruction, a first feedforward control is obtained according to the inlet concentration compensation algorithm; Considering the complex working conditions and dynamic characteristics of thermal power generating units, the first feedforward control based on the denitrification inlet concentration prediction instruction is added. The load instruction and coal feed instruction of the thermal power generating unit are used as input instructions, and the denitrification inlet concentration prediction instruction is obtained through the main loop controller to judge the changing trend of the denitrification inlet concentration in advance; when the denitrification inlet concentration has not changed significantly and the instrument calibration measurement data is inaccurate, the signal for adjusting the ammonia injection regulating valve action is output in advance to reduce the fluctuation of the denitrification outlet concentration; the inlet concentration compensation algorithm is set by setting the delay constant to adapt to the working conditions , in order to suppress the sudden rise of denitrification outlet concentration. The formula of the inlet concentration compensation algorithm is: ; Where, is the feedforward channel compensation value based on the predicted instruction, The delay constant is used to adapt to the working conditions, indicating the delay time from judging the concentration change trend of the denitrification inlet to the change of the denitrification outlet concentration. is a function related to both load instruction and coal supply instruction, For coal feeding instruction, For load instructions.

[0026] Delay constant adapted to working conditions Obtained by neural network algorithm identification. This represents the lead time for the algorithm to predict changes in the denitrification outlet concentration. Since this delay varies across operating conditions, adjusting this delay constant ensures consistent algorithm performance across all conditions. When the denitrification inlet concentration prediction command surges, an ammonia injection control valve opening compensation value is calculated based on the denitrification inlet concentration prediction command.

[0027] Based on the denitration outlet concentration, a second feedforward control is obtained according to the deviation protection logic; In order to ensure the safety of environmental protection indicators, a deviation protection mechanism based on the correction feedback of the denitrification outlet concentration is designed as the second feedforward control. By real-time monitoring of the differential value of the denitrification outlet concentration, that is, the concentration change rate, and judging the change trend of the denitrification outlet concentration. When the denitrification outlet concentration is greater than the sum of the deviation protection threshold and the outlet concentration set value, and the concentration change rate is greater than the product of the sensitivity coefficient and the deviation protection threshold, the emergency control strategy is triggered, that is, a signal for quickly adjusting the opening of the ammonia injection regulating valve is output to prevent the denitrification outlet concentration from exceeding the standard. In some embodiments, based on the denitrification outlet concentration, the steps of obtaining the second feedforward control according to the deviation protection logic are: Obtain sensitivity coefficient, deviation protection threshold and outlet concentration setting value; The concentration change rate is obtained according to the denitrification outlet concentration; When the denitrification outlet concentration is greater than the sum of the deviation protection threshold and the outlet concentration set value, and the concentration change rate is greater than the product of the sensitivity coefficient and the deviation protection threshold, the emergency control strategy is triggered; If not, the emergency control strategy is not triggered; The deviation protection logic is: ; Where, is the concentration change rate, is the sensitivity coefficient, is the deviation protection threshold, is the denitrification outlet concentration, is the outlet concentration setting value, For emergency control strategies; A historical optimal ammonia injection amount is obtained based on the second historical data set; in some embodiments, the second historical data set is historical ammonia injection data, and the second historical data set includes a set of system operating parameters of the thermal power generator set under different operating conditions, and historical transient ammonia injection amounts under different operating conditions.

[0028] The third feedforward control is obtained based on the current ammonia injection amount and the historical optimal ammonia injection amount; in order to improve the economic efficiency of the thermal power generation unit system, an economic feedforward control based on historical ammonia injection data is added. By analyzing the second historical data set, the historical optimal ammonia injection amount is calculated. In real-time control, the current amount of ammonia sprayed Compared with the historical optimal ammonia injection rate For comparison, the error ( ) is weighted and output as a correction signal to avoid economic losses caused by excessive or insufficient ammonia injection. In some embodiments, the correction formula for the third feedforward control is: ; Where, is the correction signal, is the economic feedforward ammonia injection amount based on historical ammonia injection data, is the economic feedforward gain coefficient, This is the best ammonia injection amount in history. is the current ammonia injection amount.

[0029] A feedforward master controller is obtained according to the ammonia injection amount prediction instruction, the first feedforward control, the second feedforward control, and the third feedforward control. In some embodiments, the feedforward master controller is a weighted superposition of the output signal of the first feedforward control, the output signal of the second feedforward control, and the output signal of the third feedforward control. The weighted superposition formula is: ; Where, is the final output ammonia injection amount, 、 、 、 are adjustable weighting coefficients, is the ammonia injection amount prediction instruction, is the feedforward channel compensation value based on the predicted instruction, represents the output signal of the first feedforward control, The amount of ammonia injection for deviation protection based on the correction feedback of the denitrification outlet concentration is represents the output signal of the second feedforward control, The economic feedforward ammonia injection amount based on historical ammonia injection data is: represents the output signal of the third feedforward control; The feedforward loop of the thermal power generator set is controlled according to the feedforward master controller and the current ammonia injection amount is updated; in some embodiments, the feedforward master controller is used to control the ammonia injection regulating valve.

[0030] The output signals of the first, second, and third feedforward control modes are weighted and superimposed, ultimately acting on the ammonia injection regulating valve. By rationally assigning the weights of each control mode, the thermal power generator system maintains excellent control quality under all operating conditions, achieving an optimal balance between denitrification efficiency and economic efficiency, meeting environmental emission standards, and reducing operating costs.

[0031] Calculation example: The full-operation denitrification system control method based on generalized predictive control and intelligent feedforward proposed in this disclosure was put into use in a power plant. The control target was to ensure that the hourly average NOx concentration of the thermal power generating units met the environmental protection requirement of 30 mg / m 3 At the same time, ensure a certain degree of economy.

[0032] The hourly average NOx concentration of thermal power generation units before commissioning was 15.7 mg / m 3 The dynamic fluctuation range of NOx concentration is about 50mg / m 3 After the unit was put into operation, the NOx concentration was 25.6 mg / m 3 , the dynamic fluctuation range is about 20 mg / m 3 , significantly reducing the amount of ammonia injection, ensuring the economic efficiency of the power plant while meeting environmental friendliness.

[0033] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Simple modifications or equivalent substitutions of the technical solution of the present invention by ordinary technicians in this field do not deviate from the essence and scope of the technical solution of the present invention.

Claims

1. A full-condition denitrification system control method based on generalized predictive control and intelligent feedforward, characterized in that: The following steps are involved: Based on the thermal power generating units, historical data sets, current input instructions and current ammonia injection amount are respectively obtained; Using a neural network algorithm, a transfer function model is obtained based on the historical data set; The transfer function model is input into the generalized predictive controller to obtain the main loop controller; the optimization objective function of the main loop controller is: ； In the formula, To optimize the objective function, is the set value of denitrification outlet concentration, For time, is the denitrification outlet concentration, is the weight coefficient, is the change in the control variable; Based on the main loop controller, a prediction instruction is obtained according to the current input instruction; the prediction instruction at least includes a denitration inlet concentration prediction instruction and an ammonia injection amount prediction instruction; Control the main circuit of the thermal power generating set according to the prediction instruction to at least obtain current output data, wherein the current output data at least includes a denitration outlet concentration; Based on the denitration inlet concentration prediction instruction, a first feedforward control is obtained according to an inlet concentration compensation algorithm; Based on the denitration outlet concentration, obtaining a second feedforward control according to a deviation protection logic; Obtaining a historical optimal ammonia injection amount according to the historical data set; Obtaining a third feedforward control according to the current ammonia injection amount and the historical optimal ammonia injection amount; Obtaining a feedforward master controller according to the ammonia injection amount prediction instruction, the first feedforward control, the second feedforward control and the third feedforward control; The feedforward loop of the thermal power generator set is controlled according to the feedforward master controller and the current ammonia injection amount is updated.

2. The full-operating-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 1 is characterized in that: The historical data set includes a first historical data set and a second historical data set; The first historical data set is used to obtain the transfer function model using a neural network algorithm; The second historical data set is used to obtain the historical optimal ammonia injection amount.

3. The full-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 2 is characterized in that: The first historical data set includes a set of historical input data and a set of historical output data; The historical input data includes historical load, historical coal feed rate, historical ammonia injection flow rate, historical primary air door opening and historical secondary air door opening; The historical output data includes historical denitrification inlet concentration and historical denitrification outlet concentration.

4. The full-operating-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 3 is characterized in that: The current input command includes a load command and a coal supply quantity command.

5. The full-operating-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 4 is characterized in that: The formula of the inlet concentration compensation algorithm is: ； In the formula, is the feedforward channel compensation value based on the predicted instruction, It is the delay time from judging the concentration change trend of denitrification inlet to the change of denitrification outlet concentration. is a function related to both load command and coal supply command. For coal feeding quantity instruction, For load instructions.

6. The full-operating-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 5 is characterized in that: Based on the denitration outlet concentration, the steps of obtaining the second feedforward control according to the deviation protection logic are: Obtain sensitivity coefficient, deviation protection threshold and outlet concentration setting value; Obtaining a concentration change rate according to the denitration outlet concentration; Determine whether a first condition is met according to the denitration outlet concentration, the deviation protection threshold and the outlet concentration setting value; the first condition is that the denitration outlet concentration is greater than the sum of the deviation protection threshold and the outlet concentration setting value; Determining whether a second condition is met is obtained according to the concentration change rate, the sensitivity coefficient and the deviation protection threshold; the second condition is that the concentration change rate is greater than the product of the sensitivity coefficient and the deviation protection threshold; When both the first condition and the second condition are met, triggering an emergency control strategy; When the first condition is not met or the second condition is not met, the emergency control strategy is not triggered.

7. The full-operating-condition denitration system control method based on generalized predictive control and intelligent feedforward according to claim 6 is characterized in that: The correction formula of the third feedforward control is: ； In the formula, To correct the signal, is the economic feedforward gain coefficient, The best ammonia injection amount in history. is the current ammonia injection amount.

8. The method for controlling a denitration system under all operating conditions based on generalized predictive control and intelligent feedforward according to claim 1 is characterized in that: The overall feedforward controller is a weighted superposition of the output signal of the first feedforward control, the output signal of the second feedforward control, and the output signal of the third feedforward control.

9. The method for controlling a denitration system under all operating conditions based on generalized predictive control and intelligent feedforward according to claim 2 is characterized in that: The second historical data set includes a set of system operating parameters of the thermal power generating set under different operating conditions, and historical transient ammonia injection amounts under different operating conditions.

10. The method for controlling a denitration system under all operating conditions based on generalized predictive control and intelligent feedforward according to claim 1, characterized in that: The feedforward master controller is used to control the ammonia injection regulating valve.

Images