A preheating control method for tower receivers based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning

Abstract

This invention discloses a preheating control method for tower receivers based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning. The method determines the equipment parameters, operating settings parameters, and operating measurement parameters of a multi-tube exposed tower receiver. Based on the measured wind speed and direction, and according to the surface convective heat transfer correlation of the receiver considering wind direction, the surface convective heat dissipation coefficient of different tube panels of the receiver is calculated, thereby obtaining the basic heat dissipation power per unit area. ； Based on the measured average temperatures of the light-facing and back-facing surfaces of each tube screen, a collaborative rolling prediction is performed using the trained thermal stress prediction model and temperature prediction model to dynamically determine the maximum additional net heat flux density that meets safety constraints. The sum of the basic heat dissipation power per unit area of ​​different tube screens of the receiver and the maximum available additional net heat flux density obtained by the decision is taken as the total input heat flux density of the tube screen. This invention can overcome the shortcomings of existing methods, such as inapplicability to complex working conditions, slow preheating speed, and uneven preheating.

Description

Technical Field

[0001] This invention belongs to the field of solar thermal power generation technology, and specifically relates to a preheating control method for tower receivers based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning. Background Technology

[0002] Tower-type concentrating solar power generation technology is a cutting-edge direction in concentrating solar power generation. Its receiver plays a core role in converting solar radiation reflected by the mirror field into molten salt heat energy, which directly determines the system's energy conversion efficiency and operational safety.

[0003] Most current mainstream multi-tube exposed receivers use solar salt as the working fluid with a freezing point as high as 493K. To prevent the molten salt inside the receiver from solidifying during nighttime shutdown, it needs to be emptied at night and re-injected the following day upon startup. Before injection, the tube walls need to be heated using a mirror field to prevent the injected molten salt from freezing and blocking the circuit—this is the preheating process described in this patent. During the preheating stage, convective heat loss is the main factor contributing to the receiver's energy loss. Under high wind speed conditions, the temperature distribution imbalance between the windward and leeward sides due to differences in convective heat loss causes localized temperature lag, significantly affecting the preheating time and success rate. Therefore, considering the influence of wind speed and direction on the heat transfer characteristics of the receiver during the preheating process is of significant engineering importance.

[0004] Current research on preheating control for multi-tube exposed receivers suffers from significant shortcomings, with mainstream methods exhibiting fundamental limitations. Taking the Vant-Hull algorithm used in the Solar II power plant as an example, it relies solely on the temperature of the tube screen's back surface and heating duration to set the heat flux. This heat flux data is derived from engineering experience and lacks physical mechanism support. In other words, current methods struggle to address the dynamic response characteristics of convective heat dissipation to wind speed / direction during the preheating stage, resulting in control strategies being explicitly limited to low-wind-speed conditions. Simultaneously, because thermal stress cannot be measured in real time, existing technologies use temperature rise rates entirely dependent on engineering experience as limiting standards, making it difficult to accurately characterize the true safety boundaries of the equipment. Solar thermal power plants employing similar experience-based judgment models cannot optimize preheating efficiency and, more importantly, cannot achieve uniform preheating, making it difficult to address preheating failures caused by extreme high-wind-speed conditions. Summary of the Invention

[0005] In order to overcome the shortcomings of the existing technology and address the deficiencies of the existing tower solar thermal power plant receiver preheating control methods, the present invention aims to provide a tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning, so as to guide the receiver to preheat safely and efficiently under various operating conditions.

[0006] The control approach of this invention is to accurately obtain the basic heat dissipation of different tube screens under high wind speed conditions through theoretical calculation, and to quickly obtain the predicted value of the maximum equivalent thermal stress on the surface of the receiver using machine learning. This allows for the reasonable input heat flow of any tube screen of the receiver during the preheating process, thus achieving safe and rapid preheating of the receiver of the tower solar thermal power plant under complex conditions such as high winds. The calculation method of this invention mainly includes four key points: (1) calculation of convective heat transfer coefficient based on wind direction; (2) a heat flux density hierarchical decision mechanism that decomposes the heat flux density of the tube screen into the basic heat dissipation power and the additional net heat flux density; (3) collaborative rolling prediction of temperature prediction module and stress prediction model; and (4) theoretical-machine learning collaborative preheating control.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A preheating control method for a tower receiver based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning includes the following steps: Step 1: Determine the equipment parameters, operating settings parameters, and operating measurement parameters of the multi-tube exposed tower receiver; based on the equipment parameters of the multi-tube exposed tower receiver, obtain the dataset through numerical simulation or experimental methods, and train the temperature prediction model and the thermal stress prediction model. Step 2: Based on the measurements t Average temperature of each tube screen's sun-facing surface at any given time Average temperature of backlight surface Through dual temperature thresholds Determine the preheating termination condition. If not, determine to continue preheating and proceed to step 3; otherwise, end preheating. i Number the tube screen; Step 3: Based on the measured wind speed v The surface convective heat transfer coefficient of each tube panel of the receiver is calculated based on the correlation formula considering the wind direction. h i Combined with the measured average temperature of the sun-facing surface of each tube screen Calculate the basic heat dissipation power per unit area of ​​each tube panel. Q i,base ; Step 4: Employ machine learning algorithms to achieve collaborative rolling prediction between the thermal stress prediction model and the temperature prediction model, dynamically determining the maximum additional net heat flux density that can be implemented to meet safety constraints. Q i,add ; Step 5: Based on the basic heat dissipation power per unit area of ​​each tube panel Q i,base The maximum additional net heat flux density that can be invested with the decision. Q i,add Determine the total input heat flux density of each tube panel. Qi .

[0008] In one embodiment, the equipment parameters include: heat absorption tower height. H tower Heat absorber diameter D rec Heat absorber height H rec heat absorber diameter D tube Number of tube screens N panel Infrared emissivity of pipe wall e And the heat absorber material (used to set conditions for simulation, thus providing a database for machine learning).

[0009] In one embodiment, the operating settings parameters include: the maximum input heat flux during the preheating process. Q max Minimum temperature at the end of preheating T min Preheating end threshold maximum temperature T max The maximum allowable thermal stress during the preheating process s max Measurement data interval Δ t During the preheating process of the receiver, every Δ t The time measurement parameters for a single run include: ambient temperature. T amb The tower-type solar thermal power plant's receiver height and wind speed v and the average temperature of the light-facing surface of each tube panel on the surface of the receiver. Average temperature of backlight surface relative azimuth angle of the surface α i (The angle between the normal of the tube screen and the wind direction).

[0010] In one embodiment, the input to the temperature prediction model is the additional net heat flux density. Q i,add , t Average temperature of each tube screen's sun-facing surface at any given time Average temperature of backlight surface The output is Δ t Average temperature of the back of the sunlit surface Average temperature of backlight surface The input to the thermal stress prediction model is Δ t Average temperature of the back of the sunlit surface Average temperature of backlight surface The output is the Δ of the tube screen. t Maximum equivalent thermal stress after .

[0011] In one embodiment, step 3, the surface convection heat dissipation coefficient h i The calculation process is as follows, obtained by considering the convective heat transfer correlation based on wind direction: (1) In the formula, Nu i The mixed convection Nusselt number refers to the number of different tube screen surfaces of the receiver, which is calculated using equation (2); l The thermal conductivity of air; (2) In the formula, Nu i,fc The Nusselt number refers to the forced convection number on different tube surfaces of the receiver. Nu nc The Nusselt number refers to the natural convection of the rough surface of the heat absorber tube. Nu i,fc and Nu nc Calculate using equations (3) and (4) respectively: (3) (4) In the formula, α i The relative azimuth angle of different tube panels in the receiver is the angle between the normal of the analyzed receiver tube panel and the incoming airflow direction, 0 ≤ α i ≤ π , α cr It is 0.5625 π , Re Using equation (5), the applicable range of the forced convection heat transfer correlation is 2.9 × 10⁻⁶. 6 ≤ Re ≤2.5×10 7 0.9≤ H rec / D rec ≤1.2, 33.4 mm≤ D tube ≤51.0 mm, Nu H c is the natural convection Nusselt number of a cylinder (referring to a heat absorber) under smooth conditions. 10 c 11 c 12 b 10 b 11 b 12 c 20c 21 c 22 b 20 b 21 b 22 These are the fitting parameters; possible values ​​are as follows: c 10 =-4.812、c 11 =2.686、c 12 =-1.487、b 10 =0.845, b 11 =0.020、b 12 =0、c 20 =-15.073、c 21 =8.099、c 22 =-1.271、b 20 =1.687、b 21 =-0.670、b 22 =0.120; (5) In the formula, p air density / kg·m -3 ; m Dynamic viscosity coefficient (kg·s·m) -2 .

[0012] In one embodiment, the basic heat dissipation power per unit area in step 3 Q i,base The calculation formula is as follows: (6) In the formula: Q i,con and Q i,rad These represent the convective and radiative heat dissipation power per unit area of ​​different tube panels of the receiver, respectively (W·m). -2 ; T amb Ambient temperature in K. e The infrared emissivity of the tube wall.

[0013] In one embodiment, step 4 is based on the measured average temperature of the light-facing surface of each tube screen. Average temperature of backlight surface To make an additional heat flux density decision, the following is achieved: (1) Define the maximum input heat flow in the preheating process Q max and Q i,base The difference is the additional net heat flux density. Q i,add The initial value; (2) Based on the temperature prediction model, input Q i,add The screen's power at that moment was measured. and Predict Δ t The average temperature of the sun-facing surface of the tube screen afterwards Average temperature of backlight surface ; (3) Based on the thermal stress prediction model, input the predicted temperature and Predict Δ t The maximum equivalent thermal stress of the tube screen afterwards ; (4) Determine the maximum additional net heat flux density that can be applied to the tube screen. Q i,add Does it meet the following constraints:

[0014] If not satisfied, then Q i,add Subtract Δ Q As the updated value, repeat steps (2)-(4) until the stated constraint condition is met or Q i,add Until it equals 0.

[0015] In one embodiment, to meet the accuracy requirements of the input heat flux density provided by the mirror field of an actual tower-type solar thermal power plant while ensuring the uniformity of preheating, Δ Q The value satisfies 0.5 kW / m 2 ≤Δ Q ≤2 kW / m 2 .

[0016] In one embodiment, step 5, Q i The unit is kW·m -2 The value is no greater than Q i,base and Q i,add The sum is an integer. This value is adopted to accommodate the controllability of tower-type solar thermal power plants in practical engineering applications.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. In handling complex wind disturbance conditions, by accurately quantifying the heat dissipation power of each tube screen under different incoming flow directions (step 3), the problem of local tube screen temperature being too low due to ignoring the influence of wind direction differences in traditional methods is solved, ensuring the uniformity of heat absorber preheating under strong wind conditions and achieving the required preheating temperature across the entire heat absorber area.

[0018] 2. In the global preheating process control, the traditional experience-dependent temperature rise rate limitation is broken through. By considering the convective heat transfer correlation of wind direction (step 3), the differential heat dissipation effect of wind on different tube screens is quantified. Combined with the thermal stress prediction model and temperature prediction model of machine learning (step 4), dynamic optimization is achieved. The thermal stress safety threshold is used to replace the experience temperature rise limit, so that the preheating time is shortened by more than 60%.

[0019] 3. During the execution of engineering instructions, decisions are driven by easily measurable data on the temperature and wind speed of both sides of the tube screen, and the output heat flux density is constrained to an integer value (unit: kW·m). - ²) and match the actual engineering precision (step 5) to achieve precise conversion from theoretical instructions to actual control.

[0020] In summary, this invention can accurately obtain the basic heat dissipation of different tube screens under high wind speed conditions, and quickly obtain the predicted value of the maximum equivalent thermal stress on the surface of the receiver to replace the empirical temperature rise rate as the constraint for heat flux density. This enables safe and rapid preheating of the receiver in tower solar thermal power plants under complex conditions such as high winds, and provides technical support for the high reliability operation of solar thermal power plants. Attached Figure Description

[0021] Figure 1 This is a flowchart of the coupling calculation process of the present invention.

[0022] Figure 2 A diagram visualizing the model training results.

[0023] Figure 3 The graphs show the temperature changes of the receiver tubes and screens under high wind speed (14 m / s) conditions, using the method and the Vant-Hull method respectively.

[0024] Figure 4 The image shows the temperature distribution of the receiver tubes and screens after preheating for 6 minutes using this method under high wind speed (14 m / s) conditions. Detailed Implementation

[0025] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings and examples.

[0026] The present invention provides a specific implementation of a tower-type heat absorber preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning, as described in the example.

[0027] The following is combined Figure 1 Taking the equipment parameters of the absorber in the Hami tower molten salt solar thermal power plant in Xinjiang as an example, the exemplary embodiment of the present invention will be described in detail, mainly including the following steps: Step 1: Determine the equipment parameters, operating settings parameters, and operating measurement parameters of the multi-tube exposed tower receiver. Specific values ​​are shown in Table 1, which includes the relative azimuth angles of different tube panel surfaces of the receiver.α i Calculated based on wind direction, where i The tube screen is numbered. It should be noted that, firstly, after the receiver completes the initial measurement of the operating parameters, it enters a loop. Subsequent parameters are affected by the operating decision results, so only the initial operating measurement parameters are listed; secondly, the wind speed remains constant during the simulation, taking the local annual maximum wind speed, and the ambient temperature remains constant, taking the average daily temperature on the vernal equinox.

[0028] Based on the equipment parameters of a multi-tube exposed tower receiver, numerical simulation was performed using a coupled optical-thermal-mechanical model of the receiver, resulting in a dataset of 920 data points. This dataset was then used to train temperature and stress prediction models. The coupled optical-thermal-mechanical model comprises three sub-models: an optical model simulating solar radiation propagation and absorption using the MCRT method to obtain the heat flux density provided to the receiver surface by the mirror field; a flow heat transfer model simulating energy transfer between the receiver tube wall and the tube interior using the one-dimensional lumped parameter method to obtain the temperature distribution on the receiver surface; and a receiver stress model that comprehensively considers multiple influencing factors to obtain the stress distribution on the receiver surface.

[0029] The core training process adopts a three-layer validation system: (1) Five-fold cross-validation: the temperature model is trained for 300 cycles per fold, with an early stopping mechanism (patience value = 15) to prevent overfitting; the stress model is trained for 300 cycles per fold, with an early stopping patience value = 10. (2) Full data training: both models are deeply optimized with 5000 cycles and a batch size of 16, with a custom early stopping mechanism (patience value = 12). (3) Independent testing and evaluation: the coefficient of determination (R²) is calculated based on the destandardized training set to evaluate the goodness of fit; a test set is reserved to verify the generalization ability and quantify the accuracy of the model's temperature-stress coupling prediction on unknown data. Among them, the training set: validation set: test set = 4:1:1. The training results are as follows: Figure 2 As shown, the prediction errors of both the temperature prediction model and the stress model are within 5%, and the lowest R² value of the model's predicted output on the test set is 0.988, indicating that the prediction model is reliable.

[0030] Table 1 Example System Parameters

[0031] Step 2: Based on the measured average temperature of the light-facing side and the average temperature of the back-facing side of each tube screen at that moment ( , ), through dual temperature thresholds ( Determine the preheating termination condition. If it is determined that preheating should continue, proceed to step 3. Step 3: Based on the currently measured wind speed v The convective heat transfer coefficients of different surfaces of the receiver are calculated based on the correlation formula for convective heat transfer on the receiver surface considering wind direction. hi The result is obtained by using equation (7): (7) In the formula, Nu i The mixed convection Nusselt number refers to the number of different tube screen surfaces of the receiver, which is calculated using equation (8); l is the thermal conductivity of air.

[0032] (8) In the formula, Nu i,fc The Nusselt number refers to the forced convection number on different tube surfaces of the receiver. Nu nc The Nusselt number refers to the natural convection of the rough surface of the heat absorber tube. Nu i,fc and Nu nc Calculate using equations (9) and (10) respectively: (9) (10) In the formula, α cr It is 0.5625 π , Re Using equation (11), the applicable range of the forced convection heat transfer correlation is 2.9 × 10⁻⁶. 6 ≤ Re ≤2.5×10 7 0.9≤ H rec / D rec ≤1.2, 33.4 mm≤ D tube The value of the fitting parameters is ≤51.0 mm, and the values ​​of the other fitting parameters are shown in Table 2. Nu H The Nusselt number for natural convection in a smooth cylinder (referring to the heat absorber tube in this case) is given.

[0033] (11) In the formula, p air density / kg·m -3 ; m Dynamic viscosity coefficient (kg·s·m) -2 .

[0034] Table 2. Correlation Parameters of Forced Convection Heat Transfer in Heat Absorbers c10 -4.812 c20 -15.073 c11 2.686 c21 8.099 c12 -1.487 c22 -1.271 b10 0.845 b20 1.687 b11 0.020 b21 -0.670 b12 0 b22 0.120 Based on the calculated convective heat dissipation coefficients of different surfaces of the heat absorber hi The average temperature of the sun-facing surface of different tube screens on the absorber surface was measured. Calculate the basic heat dissipation power per unit area of ​​different tube panels of the heat absorber. Q i,base The calculation formula is as follows: (12) In the formula: Q i,con , Q i,rad These represent the convective heat dissipation power and radiative heat dissipation power per unit area of ​​different tube panels of the receiver, respectively (W·m). -2 .

[0035] Step 4: Based on Figure 2 The temperature prediction model and stress prediction model of the receiver obtained by training are shown below, based on the average temperature of the light-facing surface of each tube screen obtained by measurement. Average temperature of backlight surface The additional heat flux density decision is made through the following steps: (1) Obtain the initial value of the initial additional heat flux density according to step 3. Q i,add Its value is the maximum input heat flux during the preheating process. Q max and Q i,base difference; (2) Based on the temperature prediction model, input Q i,add The screen's power at that moment was measured. and Predict Δ t The average temperature of the sun-facing surface of the tube screen afterwards Average temperature of backlight surface ; (3) Based on the thermal stress prediction model, input the predicted temperature and The maximum equivalent thermal stress of the tube screen is predicted to be 30 seconds later. ; (4) Determine the maximum additional net heat flux density that can be applied to the tube screen. Q i,add Does it satisfy the constraints shown in equation (13)? If not, then take... Q i,add = Q i,add -Δ Q Repeat steps (2)-(4) until the condition is met or Q i,add Until 0. Where, Δ Q =1 kW / m 2 .

[0036] (13) Step 5: Based on the basic heat dissipation power per unit area of ​​different tube panels of the absorber Q i,base The maximum available additional net heat flux density obtained from the decision Q i,add Determine the total input heat flux density of the tube panel. Q i Considering the controllability of existing tower solar thermal power plants, Q i The unit is kW·m -2 The value is no greater than Q i,base and Q i,add The integer sum of the numbers.

[0037] Finally, the decision results are input into the absorber's light-heat-mechanical coupling model for simulation to obtain 30-second preheating results. Steps 2-5 are then repeated based on the obtained data to finally obtain... Figure 3 The graphs shown depict the temperature changes of the receiver tubes under high wind speed (14 m / s) conditions, using the method and the Vant-Hull method respectively. Figure 4 The temperature distribution of the receiver tube screen after preheating for 6 minutes under high wind speed conditions (14 m / s) using this method is shown.

[0038] Depend on Figure 3 It is evident that even under the local maximum annual wind speed of 14 m / s, this method can achieve successful preheating within 6 minutes while ensuring safety. Its maximum thermal stress of 448 MPa is lower than the 456 MPa of the Vant-Hull method, and significantly lower than the common thermal stress limit of 580 MPa for receiver operation. Simultaneously, the receiver surface temperature remains uniform, with a high-low temperature difference not exceeding 20K, overcoming the limitation of traditional preheating methods that cannot preheat under high wind conditions due to poor preheating uniformity. Furthermore, this method significantly improves the preheating speed, reducing the preheating time by 60% compared to the shortest time of 15 minutes, compared to the normal 15-30 minutes in actual engineering projects. Figure 4 As can be seen from the temperature distribution of different tube panels after 6 minutes of preheating of the absorber under the local maximum annual wind speed of 14 m / s, all tube panels were fully preheated, with the temperature maintained between 549 K and 559 K, reaching the standard for ending preheating in just 6 minutes. Meanwhile, the temperature distribution on the outer wall of the absorber tubes shows that the lowest temperature point of the tube panel is located at the end of the absorber tube. This point is connected to the header in actual engineering applications, and the header pipes are electrically heated, therefore it will not affect the safety of the absorber.

[0039] This invention proposes a preheating control method for tower-type solar thermal power (CSP) receivers based on the synergy of theoretical heat dissipation calculation and thermal stress machine learning, combining theoretical research with machine learning. This invention enables safe and rapid preheating of CSP receivers under complex conditions such as high winds, providing technical support for the high-reliability operation of CSP plants.

Claims

1. A preheating control method for a tower-type receiver based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning, characterized in that, Includes the following steps: Step 1: Determine the equipment parameters, operating settings parameters, and operating measurement parameters of the multi-tube exposed tower receiver; based on the equipment parameters, obtain the dataset through numerical simulation or experimental methods, and train the temperature prediction model and the thermal stress prediction model. Step 2: Based on the measurements t Average temperature of each tube screen's sun-facing surface at any given time Average temperature of backlight surface Through dual temperature thresholds Determine the preheating termination condition. If not, determine to continue preheating and proceed to step 3; otherwise, end preheating. i Number the tube screen; Step 3: Based on the measured wind speed v The surface convective heat transfer coefficient of each tube panel of the receiver is calculated based on the correlation formula considering the wind direction. h i Combined with the measured average temperature of the sun-facing surface of each tube screen Calculate the basic heat dissipation power per unit area of ​​each tube panel. Q i,base ; Step 4: Employ machine learning algorithms to achieve collaborative rolling prediction between the thermal stress prediction model and the temperature prediction model, dynamically determining the maximum additional net heat flux density that can be implemented to meet safety constraints. Q i,add ; Step 5: Based on the basic heat dissipation power per unit area of ​​each tube panel Q i,base The maximum additional net heat flux density that can be invested with the decision. Q i,add Determine the total input heat flux density of each tube panel. Q i ; In step 3, the surface convection heat dissipation coefficient h i The calculation process is as follows, obtained by considering the convective heat transfer correlation based on wind direction: (1) In the formula, Nu i The mixed convection Nusselt number refers to the number of different tube screen surfaces of the receiver, which is calculated using equation (2); λ The thermal conductivity of air; (2) In the formula, Nu i,fc The Nusselt number refers to the forced convection number on different tube surfaces of the receiver. Nu nc The Nusselt number refers to the natural convection of the rough surface of the heat absorber tube. Nu i,fc and Nu nc Calculate using equations (3) and (4) respectively: (3) (4) In the formula, α i The relative azimuth angle of different tube panels in the receiver is the angle between the normal of the analyzed receiver tube panel and the incoming airflow direction, 0 ≤ α i ≤ π , α cr It is 0.5625 π , Re Using equation (5), the applicable range of the forced convection heat transfer correlation is 2.9 × 10⁻⁶. 6 ≤ Re ≤2.5×10 7 0.9≤ H rec / D rec ≤1.2, 33.4 mm≤ D tube ≤51.0 mm, Nu H c is the natural convection Nusselt number of the absorber tube under smooth conditions. 10 c 11 c 12 b 10 b 11 b 12 c 20 c 21 c 22 b 20 b 21 b 22 These are the fitting parameters; (5) In the formula, ρ air density / kg·m -3 ; μ Dynamic viscosity coefficient (kg·s·m) -2 .

2. The tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, The equipment parameters include: heat absorption tower height. H tower Heat absorber diameter D rec Heat absorber height H rec heat absorber diameter D tube Number of tube screens N panel Infrared emissivity of pipe wall ε and heat absorber material; The operating settings parameters include: maximum heat flux input during the preheating process. Q max Minimum temperature at the end of preheating T min Preheating end threshold maximum temperature T max The maximum allowable thermal stress during the preheating process σ max Measurement data interval Δ t ; During the preheating process of the absorber, every Δ t The time measurement of parameters for a single run includes: ambient temperature. T amb The tower-type heat absorber has a high wind speed. v and the average temperature of the light-facing surface of each tube panel on the surface of the receiver. Average temperature of backlight surface Relative azimuth angle of the surface α i .

3. The tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, The input to the temperature prediction model is the additional net heat flux density. Q i,add , t Average temperature of each tube screen's sun-facing surface at any given time Average temperature of backlight surface The output is Δ t Average temperature of the back of the sunlit surface Average temperature of backlight surface The input to the thermal stress prediction model is Δ t Average temperature of the back of the sunlit surface Average temperature of backlight surface The output is the tube screen Δ t Maximum equivalent thermal stress after .

4. The tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, The fitting parameters are set to the following values: c 10 =-4.812、c 11 =2.686、c 12 =-1.487、b 10 =0.845、b 11 =0.020、b 12 =0、c 20 =-15.073、c 21 =8.099、c 22 =-1.271、b 20 =1.687、b 21 =-0.670、b 22 =0.120。 5. The tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, The basic heat dissipation power per unit area in step 3 Q i,base The calculation formula is as follows: (6) In the formula: Q i,con and Q i,rad These represent the convective and radiative heat dissipation power per unit area of ​​different tube panels in the receiver, in W·m². -2 ; T amb Ambient temperature, in Kelvin (K). ε The infrared emissivity of the tube wall.

6. The preheating control method for a tower receiver based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, Step 4 is based on the measured average temperature of the sun-facing surface of each tube screen. Average temperature of backlight surface To make an additional heat flux density decision, the following is achieved: (1) Define the maximum input heat flow in the preheating process Q max and Q i,base The difference is the additional net heat flux density. Q i,add Initial value; (2) Based on the temperature prediction model, input Q i,add The screen at time t was measured. and Predict Δ t The average temperature of the sun-facing surface of the tube screen afterwards Average temperature of backlight surface ; (3) Based on the thermal stress prediction model, input the predicted temperature and Predict Δ t The maximum equivalent thermal stress of the tube screen afterwards ; (4) Determine the maximum additional net heat flux density that can be applied to the tube screen. Q i,add Does it meet the following constraints: If not satisfied, then Q i,add Subtract Δ Q As the updated value, repeat steps (2)-(4) until the stated constraint condition is met or Q i,add Until it equals 0.

7. The tower receiver preheating control method based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, To meet the accuracy requirements of the input heat flux density provided by the mirror field in a real tower solar thermal power plant while ensuring the uniformity of preheating, Δ Q The value satisfies 0.5 kW / m 2 ≤Δ Q ≤2 kW / m 2 .

8. The preheating control method for a tower receiver based on the synergy of heat dissipation theoretical calculation and thermal stress machine learning as described in claim 1, characterized in that, Step 5, Q i The unit is kW·m -2 The value is no greater than Q i,base and Q i,add The integer sum of the numbers.

Images