Boiler peak shaving system control optimization method and device based on deep learning

Through deep learning models, the real-time parameters of boilers and soda systems are predicted, and the primary, secondary and water-cooled wall pipe valves are controlled, which solves the problems of unstable combustion and soda flow during the boiler depth peak regulating process, and realizes timely and accurate boiler adjustment Peak and safety improvement.

CN119084924APending Publication Date: 2024-12-06JIANGSU GUOHUACHENJIAGANG POWER GENERATION CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411452189.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-17
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Instability in the combustion and unstable soda flow of boilers during deep peak shaving lead to economic losses and safety hazards. The existing solutions are usually to wait until the problem is serious before dealing with it, making it difficult to avoid major problems in a timely manner.

Method used

Using deep learning-based boiler peak regulating system control method, the real-time parameters of boilers and soda systems are predicted through machine learning models, thermal efficiency and thermal deviation coefficients are calculated, primary, secondary and water-cooled wall pipe valves are controlled in real time, ensuring combustion and steam Stability of flow.

Benefits of technology

It realizes timely and accurate control of the boiler under deep peak shaving conditions, improves combustion efficiency, reduces water-cooled wall heat deviation, avoids pipe burst accidents, and improves system stability and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119084924A_ABST
    Figure CN119084924A_ABST
Patent Text Reader

Abstract

The invention provides a boiler peak regulation system control optimization method and device based on deep learning, and the method is characterized in that the method comprises the following steps: S1, inputting the real-time data of a boiler into a machine learning model, and obtaining a boiler side prediction parameter and a steam-water side prediction parameter; s2, the heat efficiency of the boiler is calculated according to the boiler side prediction parameters; s3, calculating according to the steam-water measurement and prediction parameters to obtain a thermal deviation coefficient of each water-cooled wall tube in the steam-water system; s4, valves corresponding to primary air and secondary air of the boiler are controlled to be opened or closed according to the heat efficiency; and S5, the opening degree of an inlet valve of each water cooling wall tube is controlled according to the water cooling wall thermal deviation coefficient. In short, the method can realize timely and accurate deep peak regulation of the boiler.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a boiler peak-shaving system control optimization method, and in particular to a boiler peak-shaving system control optimization method and device based on deep learning. Background Art

[0002] my country's energy endowment of rich coal, poor oil and little gas determines that my country's energy structure is dominated by coal. However, the greenhouse effect caused by the coal burning process is an important challenge that all mankind must face. In order to achieve carbon peak and carbon neutrality as soon as possible, my country has gradually shifted from a coal-based power generation pattern to a new energy-based power generation pattern. However, due to the increasing energy demand in my country, the cyclical and instability of new energy and the development of energy storage technology lagging behind demand, it is also necessary to retain the functions of coal-fired boiler units to play the role of basic load and deep peak regulation.

[0003] Deep peak regulation of boilers has caused many problems. On the furnace side, when the load is reduced, the combustion temperature in the furnace decreases, resulting in unstable combustion and even flameout, causing huge economic losses. On the steam-water side, due to the instability of the flame under low load, the mass flow rate of steam and water in the water-cooled wall tubes varies greatly, causing instability in the steam-water flow, and excessive local stress can easily lead to tube burst accidents.

[0004] In order to solve the problem of combustion instability, the method of in-furnace air distribution is often used; in order to solve the problem of steam-water flow instability, the valve opening of the problem pipeline is often changed; however, since the problem is often not dealt with until serious economic and safety consequences have already occurred, how to solve the problem in a timely manner when it appears and avoid major problems is a difficult problem in achieving boiler peak regulation. Summary of the invention

[0005] The present invention is made to solve the above-mentioned problems, and its purpose is to provide a boiler peak-shaving system control optimization method and device based on deep learning.

[0006] The present invention provides a control optimization method for a boiler peak-shaving system based on deep learning, which is used to control the boiler and the steam-water system through the real-time data of the boiler, so that the combustion and the steam-water flow are stable under the deep peak-shaving conditions, and has the following characteristics, including the following steps: step S1, inputting the real-time data of the boiler into a machine learning model to obtain boiler side prediction parameters and steam-water measurement prediction parameters; step S2, calculating the thermal efficiency of the boiler according to the boiler side prediction parameters; step S3, calculating the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the steam-water measurement prediction parameters; step S4, controlling the valves corresponding to the primary air and secondary air of the boiler to open or close according to the thermal efficiency; step S5, controlling the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient, wherein the real-time data of the boiler includes coal supply, air supply, water supply, water supply temperature, furnace outlet temperature and oxygen content, the boiler side prediction parameters include flue gas velocity, flue gas temperature and flue gas components, and the steam-water measurement prediction parameters include mass flow rate, temperature, pressure and steam content.

[0007] In the boiler peak-shaving system control optimization method based on deep learning provided by the present invention, it can also have the following characteristics: wherein, the machine learning model is constructed using a physical information neural network algorithm, and the machine learning model is trained based on existing boiler real-time data and corresponding combustion simulation results and steam-water simulation results, and the combustion simulation results and steam-water simulation results are obtained by inputting input parameters into a numerical model for numerical simulation, and the input parameters include fuel characteristics, boiler structure, and operating conditions.

[0008] In the deep learning-based boiler peak-shaving system control optimization method provided by the present invention, it can also have the following characteristics: wherein, the fuel characteristics include data corresponding to elemental analysis, industrial analysis and calorific value, the boiler structure includes data corresponding to the furnace, the horizontal flue and the heating surface arranged in the horizontal flue, and the operating conditions include 20%, 40%, 60%, 80%, and 100% of the rated load.

[0009] In the boiler peak-shaving system control optimization method based on deep learning provided by the present invention, it can also have the following characteristics: wherein, in the numerical model, the combustion process simulation adopts a non-premixed combustion model, the turbulence simulation adopts a k-ε model, the coal powder pyrolysis simulation adopts a two-equation model, the coke combustion simulation adopts a diffusion dynamics model, and the radiation adopts a P1 model.

[0010] The boiler peak-shaving system control optimization method based on deep learning provided by the present invention may also have the following characteristics: wherein, in the numerical model, steam-water simulation is performed by solving the mass conservation, momentum conservation, energy conservation and state equation to obtain the steam-water simulation result.

[0011] In the boiler peak load regulation system control optimization method based on deep learning provided by the present invention, it can also have the following characteristics: wherein the calculation expression of thermal efficiency is: η=100-q A , Where η is the thermal efficiency, q A is the flue gas heat loss, T s is the flue gas temperature, T a is the air temperature, O 2 ' is the oxygen concentration, A 2 and B are preset empirical parameters.

[0012] In the boiler peak load regulation system control optimization method based on deep learning provided by the present invention, it can also have the following characteristics: wherein the calculation expression of the thermal deviation coefficient is: In the formula is the thermal deviation coefficient of the dth water-cooled wall tube, q d is the heat load on the outer wall of the dth water-cooled wall tube, H d is the heating area of ​​the dth water-cooled wall tube, G d is the working fluid flow rate of the dth water-cooled wall tube, q 0 is the average heat load on the outer wall of all water-cooled wall tubes, H 0 is the average heating area of ​​all water-cooled wall tubes, G 0 is the average working fluid flow rate of all water-cooled wall tubes.

[0013] In the deep learning-based boiler peak-shaving system control optimization method provided by the present invention, it can also have the following characteristics: wherein, in step S4, when the thermal efficiency is below 40%, the valves corresponding to the primary air at the bottom layer and the nearest upper and lower secondary air are controlled to be opened, and the valves corresponding to all other primary air and secondary air are closed; when the thermal efficiency is between 40% and 80%, the valves corresponding to the primary air at the top layer and the nearest upper and lower secondary air are controlled to be closed, and the valves corresponding to all other primary air and secondary air are opened; when the thermal efficiency is between 80% and 90%, the valves corresponding to all primary air and secondary air are controlled to be opened.

[0014] In the deep learning-based boiler peak-shaving system control optimization method provided by the present invention, it can also have the following characteristics: wherein, in step S5, when the water-cooled wall thermal deviation coefficient corresponding to the water-cooled wall tube is greater than or equal to 10%, the opening degree of the inlet valve corresponding to the water-cooled wall tube is increased by 5%; when the water-cooled wall thermal deviation coefficient corresponding to the water-cooled wall tube is less than or equal to 10%, the opening degree of the inlet valve corresponding to the water-cooled wall tube is reduced by 5%.

[0015] The present invention also provides a boiler peak-shaving system control optimization device based on deep learning, which is used to control the boiler and the steam-water system through the real-time data of the boiler, so that the combustion and the steam-water flow are stable under the deep peak-shaving conditions, and has the following characteristics, including: a prediction module, which includes a machine learning model, which is used to input the real-time data of the boiler into the machine learning model to obtain the boiler side prediction parameters and the steam-water measurement prediction parameters; a thermal efficiency calculation module, which is used to calculate the thermal efficiency of the boiler according to the boiler side prediction parameters; a thermal deviation calculation module, which is used to calculate the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the steam-water measurement prediction parameters; a primary and secondary air control module, which is used to control the opening or closing of the valves corresponding to the primary air and secondary air of the boiler according to the thermal efficiency; a water-cooled wall tube control module, which is used to control the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient, wherein the real-time data of the boiler includes coal supply, air supply, water supply, water supply temperature, furnace outlet temperature and oxygen content, the boiler side prediction parameters include flue gas velocity, flue gas temperature and flue gas components, and the steam-water measurement prediction parameters include mass flow rate, temperature, pressure and steam content.

[0016] Functions and Effects of the Invention

[0017] According to the boiler peak-shaving system control optimization method and device based on deep learning involved in the present invention, because, through combustion simulation and steam-water simulation, accurate combustion simulation results and steam-water simulation results corresponding to the parameters that cannot be measured in the actual boiler are obtained according to the real-time data of the boiler, and then the historical real-time data of the boiler, the combustion simulation results and the steam-water simulation results are used to train the machine learning model, so that the machine learning model can accurately predict the boiler parameters according to the real-time data of the boiler, and calculate the thermal efficiency and thermal deviation coefficient according to the predicted results, so as to control the primary and secondary air valves and the water-cooled wall tube valves to achieve deep peak-shaving of the boiler. Therefore, the boiler peak-shaving system control optimization method and device based on deep learning of the present invention can achieve timely and accurate deep peak-shaving of the boiler. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a block diagram of a boiler peak load regulation device in an embodiment of the present invention;

[0019] Figure 2 It is a structural schematic diagram of the primary and secondary air ducts in an embodiment of the present invention;

[0020] Figure 3 is a schematic diagram of a simulated soda flow network system in an embodiment of the present invention;

[0021] Figure 4 It is a flow chart of a boiler peak-shaving system control optimization method based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the following embodiments and the accompanying drawings specifically illustrate the boiler peak-shaving system control optimization method and device based on deep learning of the present invention.

[0023] In this embodiment, a boiler peak-shaving system control optimization device based on deep learning is provided, hereinafter referred to as the boiler peak-shaving device, which is used to control the boiler and the steam-water system through the real-time data of the boiler to stabilize the combustion and the steam-water flow under deep peak-shaving conditions.

[0024] Among them, the real-time data of the boiler include coal supply, air supply, water supply, water supply temperature, furnace outlet temperature and oxygen content.

[0025] Figure 1 is a block diagram of a boiler peak-shaving device in an embodiment of the present invention.

[0026] like Figure 1 As shown, the boiler peak-shaving device 100 includes a prediction module 10, a thermal efficiency calculation module 20, a thermal deviation calculation module 30, a primary and secondary air control module 40, a water-cooled wall tube control module 50 and a general control module 60 for controlling the operation of the above modules.

[0027] The prediction module 10 includes a machine learning model, which is used to input the real-time data of the boiler into the machine learning model to obtain boiler side prediction parameters and steam-water test prediction parameters.

[0028] Among them, the prediction parameters on the boiler side include flue gas velocity, flue gas temperature and flue gas components, and the prediction parameters on the steam-water side include mass flow rate, temperature, pressure and steam content.

[0029] Among them, the machine learning model is constructed using a physical information neural network algorithm. The model is a physical-informed neural network, or PINNs. The steps of the model include: data acquisition, selection and construction of physical models, training, verification and testing of PINNs models, and model predictive control. In order to make the results more consistent with the laws of physics, this study incorporates the continuity equation, momentum conservation equation, and energy conservation equation into the training group of the neural network.

[0030] The machine learning model is trained based on the existing real-time boiler data and the corresponding combustion simulation results and steam-water simulation results.

[0031] The combustion simulation results and steam-water simulation results are obtained by inputting the input parameters into the numerical model for numerical simulation.

[0032] Input parameters include fuel characteristics, boiler structure and operating conditions. Fuel characteristics include elemental analysis, industrial analysis and calorific value data. Boiler structure includes furnace, horizontal flue and heating surface data arranged in the horizontal flue. Operating conditions include 20%, 40%, 60%, 80% and 100% rated load.

[0033] Figure 2 It is a schematic structural diagram of the primary and secondary air ducts in an embodiment of the present invention.

[0034] like Figure 2 As shown, primary air nozzles and secondary air nozzles are respectively arranged on both sides of the boiler, and the primary air nozzles and secondary air nozzles on each side are arranged alternately. Each primary air nozzle is controlled by a corresponding primary air valve to form a primary air duct, and each secondary air nozzle is controlled by a corresponding secondary air valve to form a secondary air duct. The furnace outlet measurement point is used to obtain the temperature and oxygen content there, which are used as the main verification parameters for combustion simulation. The operating data of the temperature and oxygen content of the furnace outlet changing with time during the operation of the boiler is used as training data for machine learning.

[0035] In this embodiment, the boiler furnace and horizontal flue are selected for modeling. According to the actual boundary conditions, that is, the opening of the secondary air damper, the secondary air duct is modeled to obtain the boundary conditions of the secondary air speed required for the combustion process. The primary air speed is a known condition. Different primary and secondary air control strategies are adopted according to different loads. When the rated load is above 80%, a strategy of fully opening the primary and secondary air is adopted; when the rated load is between 40-80%, a two-layer primary air strategy is adopted, and the top layer of primary air and the corresponding upper and lower secondary air are closed at the same time; when the rated load is below 40%, a one-layer primary air strategy is adopted to close the upper two layers of primary air and the corresponding upper and lower secondary air. Through the primary and secondary air control strategies under the above-mentioned different working conditions, combined with other parameter conditions, the combustion simulation results are obtained according to the numerical model.

[0036] In the numerical model, the non-premixed combustion model is used to simulate the combustion process, the k-ε model is used to simulate the turbulence, the two-equation model is used to simulate the pyrolysis of coal powder, the diffusion dynamics model is used to simulate the coke combustion, and the P1 model is used for radiation.

[0037] In the numerical model, the soda-water simulation is carried out by solving the mass conservation, momentum conservation, energy conservation and state equation to obtain the soda-water simulation results.

[0038] In this embodiment, the combustion simulation result provides the boundary conditions of heat load distribution for the steam-water simulation, that is, a combustion simulation calculation is first performed to obtain the combustion simulation result, and then a steam-water simulation calculation is performed based on the combustion simulation result to obtain the steam-water simulation result.

[0039] Figure 3 Schematic diagram of a simulated soda flow network system in an embodiment of the present invention.

[0040] like Figure 3 As shown in the figure, the simulated steam-water flow network system includes an economizer outlet header, a front water-cooled wall inlet header, a rear water-cooled wall inlet header, water-cooled wall tubes, and a water-cooled wall outlet mixing header. Multiple water-cooled wall tubes constitute the four furnace walls, namely the left wall, the front wall, the right wall, and the rear wall, and each water-cooled wall tube is provided with a corresponding temperature and pressure measuring point and a water-cooled wall tube valve. The temperature and pressure measuring point can only detect the temperature and pressure of the local water-cooled wall tube, and cannot fully feedback the temperature and pressure of the entire water-cooled wall tube. Therefore, according to the boundary conditions of the heat load distribution, in the steam-water simulation process, all water-cooled wall tubes are simulated in combination with the simulated steam-water flow network system, and accurate steam-water simulation results are obtained in combination with the numerical model, thereby realizing accurate prediction of the temperature and pressure of the entire water-cooled wall tube.

[0041] In this embodiment, accurate combustion simulation results and steam-water simulation results are obtained based on the real-time data of the boiler, and then the machine learning model is trained based on the combustion simulation results and the steam-water simulation results to achieve accurate prediction of the boiler side prediction parameters and the steam-water side prediction parameters.

[0042] The thermal efficiency calculation module 20 is used to calculate the thermal efficiency of the boiler according to the boiler-side prediction parameters.

[0043] The calculation expression of thermal efficiency is:

[0044] η=100-q A ,

[0045]

[0046] Where η is the thermal efficiency, q A is the flue gas heat loss, T s is the flue gas temperature, T a is the air temperature, O 2 ' is the oxygen concentration, A 2 and B are preset empirical parameters.

[0047] The values ​​of the empirical parameters of various fuels in this embodiment are shown in the following table:

[0048] Fuel Type <![CDATA[A 2 ]]> B natural gas 0.66 0.009 Oil fuel 0.68 0.007 gas 0.63 0.011 Coke oven gas 0.60 0.011 Liquefied Natural Gas 0.63 0.008

[0049] The first column in the table above is the fuel, and the second column is the empirical parameter A corresponding to each fuel. 2 , the third column is the empirical parameter B corresponding to each fuel.

[0050] The thermal deviation calculation module 30 calculates the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the predicted parameters of the steam-water measurement.

[0051] Among them, the calculation expression of thermal deviation coefficient is:

[0052]

[0053] In the formula is the thermal deviation coefficient of the dth water-cooled wall tube, q d is the heat load on the outer wall of the dth water-cooled wall tube, H d is the heating area of ​​the dth water-cooled wall tube, G d is the working fluid flow rate of the dth water-cooled wall tube, q 0 is the average heat load on the outer wall of all water-cooled wall tubes, H 0 is the average heating area of ​​all water-cooled wall tubes, G 0 is the average working fluid flow rate of all water-cooled wall tubes.

[0054] The primary and secondary air control module 40 is used to control the opening or closing of valves corresponding to the primary and secondary air of the boiler according to the thermal efficiency.

[0055] Among them, when the thermal efficiency is below 40%, the primary and secondary air control module 40 controls the valves corresponding to the primary air at the bottom layer and the nearest upper and lower secondary air to open, and all other valves corresponding to the primary air and secondary air are closed; when the thermal efficiency is between 40% and 80%, the primary and secondary air control module 40 controls the valves corresponding to the primary air at the top layer and the nearest upper and lower secondary air to close, and all other valves corresponding to the primary air and secondary air are opened; when the thermal efficiency is between 80% and 90%, the primary and secondary air control module 40 controls all valves corresponding to the primary air and secondary air to open.

[0056] The water-cooled wall tube control module 50 is used to control the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient.

[0057] Among them, when the water-cooled wall thermal deviation coefficient corresponding to the water-cooled wall tube is greater than or equal to 10%, the water-cooled wall tube control module 50 controls the opening degree of the inlet valve corresponding to the water-cooled wall tube to be increased by 5%; when the water-cooled wall thermal deviation coefficient corresponding to the water-cooled wall tube is less than or equal to 10%, the water-cooled wall tube control module 50 controls the opening degree of the inlet valve corresponding to the water-cooled wall tube to be decreased by 5%.

[0058] The master control module 60 stores a control program for controlling the operation of each module.

[0059] The following, in conjunction with the accompanying drawings, describes the process of using the boiler peak-shaving device 100 to perform a deep learning-based boiler peak-shaving system control optimization method.

[0060] Figure 4 It is a flow chart of a boiler peak-shaving system control optimization method based on deep learning in an embodiment of the present invention.

[0061] like Figure 4 As shown in the figure, the boiler peak load regulation system control optimization method based on deep learning includes the following steps:

[0062] Step S1, using the prediction module 10 to input the real-time data of the boiler into the machine learning model to obtain the boiler side prediction parameters and the steam-water test prediction parameters.

[0063] Step S2: The thermal efficiency calculation module 20 is used to calculate the thermal efficiency of the boiler according to the boiler-side prediction parameters.

[0064] Step S3, using the thermal deviation calculation module 30 to calculate the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the predicted parameters of the steam-water system.

[0065] Step S4, using the primary and secondary air control module 40 to control the opening or closing of valves corresponding to the primary and secondary air of the boiler according to the thermal efficiency.

[0066] Step S5, using the water-cooled wall tube control module 50 to control the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient.

[0067] In this embodiment, the boiler is deeply peak-regulated by the boiler peak-regulating device 100, so that the combustion efficiency can be higher than 90%, and the thermal deviation of the water-cooled wall on the steam-water side can be less than 10%.

[0068] Functions and Effects of the Embodiments

[0069] According to the deep learning-based boiler peak-shaving system control optimization method and device involved in this embodiment, through combustion simulation and steam-water simulation, accurate combustion simulation results and steam-water simulation results corresponding to parameters that cannot be measured in the actual boiler are obtained according to the real-time data of the boiler, and then the historical real-time data of the boiler, the combustion simulation results and the steam-water simulation results are used to train the machine learning model, so that the machine learning model can accurately predict the boiler parameters according to the real-time data of the boiler, and calculate the thermal efficiency and thermal deviation coefficient according to the predicted results, so as to control the primary and secondary air valves and the water-cooled wall tube valves to achieve deep peak-shaving of the boiler. In short, this method can achieve timely and accurate deep peak-shaving of the boiler.

[0070] Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A control optimization method for a boiler peak-shaving system based on deep learning, which is used to control the boiler and the steam-water system through real-time boiler data to stabilize combustion and steam-water flow under deep peak-shaving conditions, characterized in that: The following steps are involved: Step S1, inputting the real-time data of the boiler into the machine learning model to obtain boiler-side prediction parameters and steam-water test prediction parameters; Step S2, calculating the thermal efficiency of the boiler according to the boiler-side prediction parameters; Step S3, calculating the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the steam-water prediction parameters; Step S4, controlling the valves corresponding to the primary air and secondary air of the boiler to open or close according to the thermal efficiency; Step S5, controlling the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient, The real-time data of the boiler include coal supply, air supply, water supply, water supply temperature, furnace outlet temperature and oxygen content. The boiler side prediction parameters include flue gas velocity, flue gas temperature and flue gas composition. The steam-water measurement prediction parameters include mass flow rate, temperature, pressure and steam content.

2. The boiler peak load regulation system control optimization method based on deep learning according to claim 1, characterized in that: in, The machine learning model is constructed using a physical information neural network algorithm. The machine learning model is trained based on the existing real-time data of the boiler and the corresponding combustion simulation results and steam-water simulation results. The combustion simulation results and the steam-water simulation results are obtained by inputting input parameters into a numerical model for numerical simulation. The input parameters include fuel characteristics, boiler structure and operating conditions.

3. The boiler peak load regulation system control optimization method based on deep learning according to claim 2 is characterized in that: in, The fuel characteristics include data corresponding to elemental analysis, industrial analysis and calorific value, The boiler structure includes data corresponding to a furnace, a horizontal flue and a heating surface arranged in the horizontal flue, The working conditions include 20%, 40%, 60%, 80%, and 100% of the rated load.

4. The boiler peak load regulation system control optimization method based on deep learning according to claim 2 is characterized in that: in, In the numerical model, the combustion process is simulated using a non-premixed combustion model. The k-ε model is used for turbulence simulation. The two-equation model is used for the simulation of pulverized coal pyrolysis. The coke combustion simulation uses a diffusion dynamic model. The radiation adopts the P1 model.

5. The boiler peak load regulation system control optimization method based on deep learning according to claim 2, characterized in that: in, In the numerical model, the soda-water simulation is performed by solving the mass conservation, momentum conservation, energy conservation and state equation to obtain the soda-water simulation result.

6. The boiler peak load regulation system control optimization method based on deep learning according to claim 1, characterized in that: in, The calculation expression of the thermal efficiency is: n=100-q A , Where η is the thermal efficiency, q A is the flue gas heat loss, T s is the flue gas temperature, T a is the air temperature, O2' is the oxygen concentration, A2 and B are both preset empirical parameters.

7. The boiler peak load regulation system control optimization method based on deep learning according to claim 1, characterized in that: in, The calculation expression of the thermal deviation coefficient is: In the formula is the thermal deviation coefficient of the dth water-cooled wall tube, q d is the heat load on the outer wall of the dth water-cooled wall tube, H d is the heating area of ​​the dth water-cooled wall tube, G d is the working fluid flow rate of the dth water-cooled wall tube, q0 is the average heating surface outer wall heat load of all water-cooled wall tubes, H0 is the average heating area of ​​all water-cooled wall tubes, and G0 is the average working fluid flow rate of all water-cooled wall tubes.

8. The boiler peak load regulation system control optimization method based on deep learning according to claim 1, characterized in that: in, In step S4, when the thermal efficiency is below 40%, the valves corresponding to the primary air at the bottom layer and the upper and lower secondary air are controlled to be opened, and the valves corresponding to all other primary air and secondary air are closed; When the thermal efficiency is between 40% and 80%, the valves corresponding to the primary air at the top layer and the upper and lower secondary air at the nearest layer are controlled to be closed, and the valves corresponding to all other primary air and secondary air are controlled to be opened; When the thermal efficiency is between 80% and 90%, all valves corresponding to the primary air and the secondary air are controlled to open.

9. The boiler peak load regulation system control optimization method based on deep learning according to claim 1, characterized in that: in, In the step S5, when the thermal deviation coefficient of the water-cooled wall corresponding to the water-cooled wall tube is greater than or equal to 10%, the opening degree of the inlet valve corresponding to the water-cooled wall tube is increased by 5%; When the thermal deviation coefficient of the water-cooled wall corresponding to the water-cooled wall tube is less than or equal to 10%, the opening degree of the inlet valve corresponding to the water-cooled wall tube is reduced by 5%.

10. A control optimization device for a boiler peak-shaving system based on deep learning, which is used to control the boiler and the steam-water system through real-time data of the boiler to stabilize the combustion and steam-water flow under deep peak-shaving conditions, characterized in that: include: A prediction module, comprising a machine learning model, for inputting the real-time data of the boiler into the machine learning model to obtain boiler-side prediction parameters and steam-water test prediction parameters; A thermal efficiency calculation module, used for calculating the thermal efficiency of the boiler according to the boiler-side prediction parameters; A thermal deviation calculation module, used for calculating the thermal deviation coefficient of each water-cooled wall tube in the steam-water system according to the steam-water measurement prediction parameters; A primary and secondary air control module, used to control the opening or closing of valves corresponding to the primary and secondary air of the boiler according to the thermal efficiency; A water-cooled wall tube control module is used to control the opening degree of the inlet valve of each water-cooled wall tube according to the water-cooled wall thermal deviation coefficient, The real-time data of the boiler include coal supply, air supply, water supply, water supply temperature, furnace outlet temperature and oxygen content. The boiler side prediction parameters include flue gas velocity, flue gas temperature and flue gas composition. The steam-water measurement prediction parameters include mass flow rate, temperature, pressure and steam content.