Concrete dam water cooling optimization control method and system based on transfer learning

Abstract

The invention proposes a transfer learning-based concrete dam water cooling optimization control method, and the method comprises the steps: building a priori bin temperature field sample set based on a simulation calculation model, capturing a nonlinear relation between an input parameter and a priori bin temperature field calculation result through an artificial neural network, building a priori bin agent model, and carrying out the calculation of the priori bin temperature field sample set. And migrating and applying to an actual dam bin so as to remarkably reduce the number of samples required by simulation calculation. And then, on the basis of taking the maximum temperature and the cooling cost as optimization objectives and combining the cooling rate and the temperature gradient as constraint conditions, constructing a double-objective optimization model of water cooling, and performing intelligent optimization selection on a dam bin water scheme by using an NSGA-II algorithm. In the optimization process, thermal parameters are calibrated through a dynamic model updating technology, and the reliability and accuracy of an optimization result are further improved. According to the method, an accurate water cooling measure considering economical efficiency and safety can be generated in a short time, and the contradiction between short construction period and insufficient optimization efficiency in actual engineering is effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of temperature control and crack prevention in water conservancy and hydropower projects, and specifically to an optimized control method for water cooling of concrete dams based on transfer learning. Background Technique

[0002] Temperature cracks are important diseases affecting the structural safety and durability of concrete dams. At present, a number of important extra-high concrete dams are about to be built in high-altitude areas of our country. In these high-altitude areas, the daily temperature change range is large and cold waves occur frequently. For example, the HD Hydropower Station on the Lancang River is located in Lanping County, Yunnan Province. The barrage is a roller-compacted concrete gravity dam. The average annual temperature is 5°C to 16°C. However, the daily temperature difference in winter exceeds 20°C, and the extremely low temperature reaches -13°C; the main dam of the KLSK Water Conservancy Project in Xinjiang is a roller-compacted concrete gravity dam. The average annual temperature in the dam site area is 2.7°C, and the extremely low temperature is -49.8°C. Cold waves are frequent and the temperature drop is large. Each cold wave has a temperature drop of more than 10°C; the water retaining structure of the BDa Hydropower Station in Tibet is a roller-compacted concrete gravity dam. The average annual temperature observed in the dam site area is 4.3°C, the extremely high temperature is 26.1°C, and the extremely low temperature is -24.6°C. Due to the large volume of roller-compacted concrete in the dam, the internal temperature is likely to continue to rise and accumulate during the construction period due to the hydration reaction. If the temperature treatment measures are improper, the temperature difference between the inside and outside of the large-volume concrete structure will continue to increase, which will lead to the generation of temperature cracks and seriously affect the quality and service life of the large-volume concrete structure.

[0003] To prevent the generation of cracks, temperature control measures need to be taken to ensure the construction quality of the dam. At present, the commonly used temperature control and crack prevention technologies in concrete dam construction include controlling the pouring temperature, surface heat preservation, water cooling, etc. Among them, the water cooling technology is to embed a pipeline cooling system before concrete pouring and conduct water cooling after pouring, and achieve personalized temperature control goals by regulating parameters such as cooling water temperature and flow rate. Compared with other temperature control and crack prevention measures, water cooling has prominent advantages such as simple construction, high heat exchange efficiency, flexible adjustment, and low cost, and has been widely used in the construction of concrete dam structures. In order to make water cooling more effective and economical, it is necessary to plan and adjust the water cooling scheme, consider multiple factors such as cooling flow rate and cooling water temperature, and select the optimal scheme.

[0004] Due to the lag of the internal temperature change of concrete behind the adjustment of the water-cooling measures, the formulation of the water-cooling scheme for concrete dams is a complex multi-factor system optimization problem. In recent years, certain progress has been made in the research on the water-cooling control of concrete dams. For example, "Intelligent Optimization of Temperature Control Measures for Concrete Dam Pouring Bays Based on the IABAP Hybrid Algorithm" (Hydropower Energy Science, 2024) proposes taking the preset water-cooling scheme as the input condition, and making feedback adjustments according to the temperature field results obtained by simulation, so as to determine the appropriate water-cooling measures. However, the simulation calculation model has a high calculation cost, and the tight actual construction period limits the practical application of the above method. Moreover, this method only considers the factor of safety on one hand and is difficult to balance economy and safety. "A Method for Rapid Regulation of Water Cooling in the Middle and Late Stages of Concrete Dams" (CN201410311743.0) dynamically predicts the temperature response, designs temperature monitoring indicators, establishes an objective function, introduces an optimization algorithm, and selects the current optimal scheme from the feasible region of water-cooling measures to achieve real-time water-cooling regulation for several days in the future. However, it calculates the internal temperature of concrete by empirical formula by hand, with low accuracy and unable to reflect the real situation, resulting in insufficient accuracy of the optimization results. "Intelligent Optimization Method of Temperature Control Measures for Concrete Dams Based on BP Neural Network" (Hydropower Energy Science, 2017) uses a neural network model as a finite element surrogate model, with the measured maximum temperature and maximum daily temperature drop rate as inputs to optimize the water-cooling parameters. However, there are certain differences between the temperature results calculated by finite element simulation and the actual temperature of the dam body, unable to accurately reflect the real situation, resulting in insufficient accuracy of the optimization results. Summary of the Invention

[0005] Aiming at the problem that it is difficult to quickly and accurately optimize the existing water-cooling control methods for concrete dams, the present invention proposes an optimized control method for water-cooling of concrete dams based on transfer learning. Based on the simulation calculation model, the present invention constructs a temperature field sample set of the prior bay in advance, captures the non-linear relationship between the input parameters and the calculation results of the temperature field of the prior bay through an artificial neural network, establishes a prior bay surrogate model, and migrates it to the actual dam bay, significantly reducing the number of samples required for the simulation calculation model. Subsequently, a double-objective optimization model for water-cooling is constructed with the maximum temperature and cooling cost as the optimization objectives and the temperature drop rate and temperature gradient as the constraints, and the NSGA-II algorithm is used to intelligently optimize the water-cooling scheme of the dam bay. During the optimization process, the thermal parameters are calibrated by combining the dynamic model update technology, further improving the reliability and accuracy of the optimization results. Through the above method, precise water-cooling measures that balance economy and safety are generated in a short time, effectively solving the contradiction between the tight construction period and insufficient optimization efficiency in actual projects.

[0006] According to one aspect of the specification of the present invention, an optimized control method for water-cooling of concrete dams based on transfer learning is provided, including: S1. Select the prior bin size range, pouring temperature, air temperature, material parameters, and the change range of water cooling parameters according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling; S2. Adopt parametric modeling to establish a simulation model of the prior bin, obtain the calculation results of the temperature field through simulation, and use an artificial neural network surrogate model to establish the non-linear relationship between the boundary conditions and the temperature field of the prior bin, and construct a temperature field surrogate model for the prior bin; S3. Use the simulation calculation data of the temperature field of the actual pouring bin of the concrete dam as the training sample, correct the temperature field surrogate model of the prior bin by using the simulation calculation results of the actual temperature field of the dam, and transfer and learn the temperature field surrogate model of the prior bin to the actual pouring bin of the dam to obtain a temperature field surrogate model for the actual pouring bin of the concrete dam; S4. Construct a water cooling double-objective optimization mathematical model with the maximum internal temperature and water cooling cost, use the NSGA-II algorithm to solve the temperature field surrogate model of the actual pouring bin of the concrete dam for each bin to obtain the Pareto set, and select the solution with the minimum distance to the utopia point as the relative optimal solution through the equilibrium point method to obtain the relative optimal water cooling scheme for the pouring bin; S5. As the pouring process progresses and the temperature monitoring data gradually increases, adopt parameter identification technology to dynamically update the thermal parameters of the model; S6. Repeat steps S3 - S5 until the optimization of all concrete dam pouring bins is completed.

[0007] As a further technical solution, S2 further includes: S21. Obtain the geometric parameters of the dam construction; S22. Establish a typical rectangular dam bin according to the geometric parameters, and extend the dam bin downward by 1.5 times the dam height as the dam foundation; S23. Obtain the air temperature, material parameters, and water cooling parameters, use them as the input for the temperature field simulation calculation of the prior bin, and obtain the calculation results of the temperature field; S24. Use the geometric parameters, air temperature, material parameters, and water cooling parameters as the input of the artificial neural network model, and the temperature field results as the output to construct an initial artificial neural network surrogate model.

[0008] As a further technical solution, S23 further includes: Create a prior bin using parametric modeling or create a three-dimensional calculation model according to the project drawings; Determine the boundary conditions, thermodynamic parameters, and initial conditions according to the sample space generated by the concrete, and perform temperature field calculation; Integrate the output temperature field simulation results, and integrate the sample space points and the output temperature field to form a training sample set.

[0009] As a further technical solution, S3 further includes: S31. Obtain the air temperature, material parameters, and water-cooling parameters, use them as the input for the simulation calculation of the temperature field in the actual concrete placement bin, and obtain the temperature field calculation results; S32. Use the air temperature, material parameters, and water-cooling parameters as the input of the initial artificial neural network model, and the dam body temperature field results as the output, and fine-tune the neural network weights to achieve the proxy model migration from the prior bin to the actual concrete placement bin.

[0010] As a further technical solution, the S4 further includes: The double-objective optimization mathematical model for water-cooling uses the maximum temperature in the concrete placement bin as the measurement standard for safety performance, the total cooling cost as the evaluation index for economic benefits, and controls the cooling rate and temperature gradient as the constraint conditions. The double-objective optimization mathematical expression is:

[0011] In the formula: f ( x ) is the optimization objective, g ( x ) is the optimization constraint, T max is the maximum temperature in the concrete placement bin; C is the water-cooling cost; T ( t n ) is t n the temperature at time T w is the initial water temperature; T o is the cooling water temperature; Q is the total cooling water flow rate, T v is the cooling rate; T i is the temperature at the current time; T i+1 is the temperature at time i + 1; T min is the minimum temperature in the concrete placement bin; L is the horizontal distance between the maximum temperature and the minimum temperature; δ is the temperature gradient threshold; is the temperature gradient.

[0012] As a further technical solution, by using the balance point method to select the solution with the minimum distance from the utopia point as the relative optimal solution, it further includes: Define the point where both objectives reach the minimum as the utopia point, normalize each optimization objective, calculate the distance between each point on the Pareto front and the utopia point, and select the solution with the closest distance as the relative optimal solution. The expression is as follows:

[0013] wherein: x i , y i are the horizontal and vertical coordinates of each scatter point in the Pareto set; x Utopia and y Utopia are the horizontal and vertical coordinates of the utopia point.

[0014] As a further technical solution, parameter identification technology is adopted to dynamically update the thermal parameters of the model, and it further includes: Dynamically inversely calculate the thermal parameters of the concrete material according to the monitoring data generated during the pouring process, where the thermal parameters are selected as the final adiabatic temperature rise θ 0, the number of days for half of the hydration heat n , the surface heat release coefficient β one or a combination of several of them.

[0015] According to one aspect of the specification of the present invention, a migration learning-based optimized control system for water cooling of concrete dams is provided, including: The first calculation module is used to select the prior bin size range, pouring temperature, air temperature, material parameters, and the change range of water cooling parameters according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling; The second calculation module is used to adopt parametric modeling to establish a simulation model of the prior bin, and obtain the temperature field calculation result through simulation. An artificial neural network surrogate model is used to establish the nonlinear relationship between the boundary conditions of the prior bin and the temperature field, and a prior bin temperature field surrogate model is constructed; The third calculation module is used to use the temperature field simulation calculation data of the actual pouring bin of the concrete dam as the training sample, correct the prior bin temperature field surrogate model by using the actual temperature field simulation calculation result of the dam, and transfer the prior bin temperature field surrogate model to the actual pouring bin of the dam through transfer learning to obtain the actual pouring bin temperature field surrogate model of the concrete dam; The fourth calculation module is used to construct a water cooling double-objective optimization mathematical model with the internal maximum temperature and water cooling cost, use the NSGA-II algorithm to solve the actual pouring bin temperature field surrogate model of the concrete dam for each bin to obtain the Pareto set, and select the solution with the smallest distance from the utopia point as the relative optimal solution through the equilibrium point method to obtain the relative optimal water cooling scheme for the pouring bin; The fifth calculation module is used to dynamically update the thermal parameters of the model by using parameter identification technology as the pouring process progresses and the temperature monitoring data gradually increases.

[0016] According to one aspect of the specification of the present invention, there is provided an optimized control device for the water cooling of a concrete dam using transfer learning, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the optimized control method for the water cooling of a concrete dam using transfer learning as described above.

[0017] According to one aspect of the specification of the present invention, there is provided a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the optimized control method for the water cooling of a concrete dam using transfer learning as described above.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention can quickly construct a surrogate model through transfer learning, and a prior bin can be constructed before obtaining the concrete dam model. The prior bin with lower computational cost is used to provide a prior data set, and then a small amount of dam body data set is supplemented. The surrogate model of the prior bin is migrated to the actual working condition, so as to quickly obtain the calculation results of numerical simulation, solve the problem that the training process of the traditional surrogate model is complex and time-consuming, effectively reduce the difficulty of the optimization problem, and meet the requirements of tight construction period in actual engineering.

[0019] 2. The present invention establishes an optimized control objective function based on safety and economic indicators, constructs a bi-objective optimization model, and uses the NSGA-II algorithm for intelligent optimization, solving the problem that the traditional method fails to fully consider the balance between economy and safety, improving the overall efficiency of the project, and providing a new solution for the temperature control technology of concrete dams.

[0020] 3. Through the dynamic model update technology, the present invention calibrates the thermal parameters in real time during the optimization process, reduces the deviation between the numerical calculation model and the actual temperature field, realizes the accurate matching of the surrogate model and the actual working condition, and ensures the accuracy and reliability of the results. Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings used in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of an optimized control method for the water cooling of a concrete dam using transfer learning according to an embodiment of the present invention.

[0023] Figure 2 It is a comparison of the prediction accuracy between an embodiment of the present invention and a traditional surrogate model.

[0024] Figure 3 Precision comparison between the embodiment of the present invention and the traditional proxy model under different sample sets.

[0025] Figure 4 Optimization results of water cooling for a certain dam bin in the embodiment of the present invention.

[0026] Figure 5 Comparison between the calculated value and the monitored value after parameter identification in the embodiment of the present invention.

[0027] Figure 6 Optimization results of different dam sections in the embodiment of the present invention.

[0028] Figure 7 Comparison of the required time for the required sample size in the embodiment of the present invention. Detailed implementation manners

[0029] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. In addition, the technical features in each embodiment or a single embodiment provided by the present invention can be combined with each other arbitrarily to form a new technical solution, and this combination is not restricted by the order of steps and / or the mode of structural composition, but must be based on the fact that those of ordinary skill in the art can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0030] As Figure 1 shown, the embodiment of the present invention provides a migration learning-based optimization control method for water cooling of concrete dams, including the following steps: S1. Construction of sample design space: Select a reasonable range of prior bin sizes, pouring temperatures, air temperatures, material parameters, and water cooling parameter change ranges according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling.

[0031] Specifically, the Latin hypercube sampling evenly divides the value range of each component into the same intervals and performs uniform sampling within each interval.

[0032] S2. Establishment of prior bin temperature field proxy model: Adopt parametric modeling to establish a simulation model of the prior bin, obtain the temperature field calculation results through simulation, and use an artificial neural network proxy model to establish a non-linear relationship between the prior bin boundary conditions and the temperature field, and construct a prior bin temperature field proxy model.

[0033] Specifically, S2 includes the following steps: S21. Obtain geometric parameters such as the upstream width, downstream width, and longitudinal length along the river; S22. Establish a typical rectangular dam bin based on the geometric parameters, and extend the dam bin downward by 1.5 times the dam height as the dam foundation; S23. Obtain the air temperature, material parameters, and water cooling parameters, which are used as the input for the prior bin temperature field simulation calculation, and obtain the temperature field calculation results; S24. Use the geometric parameters, air temperature, material parameters, and water cooling parameters as the input of the artificial neural network model, and the temperature field results as the output to construct an initial artificial neural network surrogate model.

[0034] Specifically, the surrogate model constructs an artificial neural network through training samples to approximate the temperature field results of the simulation calculation model, significantly reducing the calculation cost.

[0035] Specifically, the temperature calculation adopts the whole-process temperature simulation calculation of concrete pouring period by finite element analysis, which consists of the following steps: (1) Create a finite element model. Create a prior bin by parametric modeling or create a three-dimensional calculation model according to the project drawings, and discretize it into a finite element model; (2) Temperature field analysis and calculation. Determine the boundary conditions, thermodynamic parameters, and initial conditions, etc. according to the sample space generated by the concrete, conduct temperature field calculation based on the finite element model, and calculate according to the basic equation of solid heat conduction:

[0036] In the formula, T is the concrete temperature; is the calculation time step; is the thermal diffusivity; is the ultimate adiabatic temperature rise of the concrete, is the pouring temperature, is the cooling water temperature, is the water cooling function.

[0037] (3) Post-processing. Integrate the output temperature field simulation results, and integrate the sample space points and the output temperature field to form a training sample set.

[0038] Specifically, the simulation calculation model can be one or a combination of a finite element model, a boundary element model, and a meshless model, and the simulation can be one or a coupling of a finite element method, a boundary element method, and a meshless method.

[0039] S3. Surrogate model transfer learning: Using the temperature field simulation calculation data of the actual concrete dam pouring bay as the training samples, the prior surrogate model is corrected by the simulation calculation results of the actual temperature field of the dam, and the prior bay surrogate model is transferred to the actual pouring bay of the dam to obtain the temperature field surrogate model of the actual pouring bay of the concrete dam.

[0040] Specifically, S3 includes the following steps: S31. Obtain the air temperature, material parameters, and water cooling parameters as the input for the temperature field simulation calculation of the actual pouring bay, and obtain the temperature field calculation results; S32. Using the air temperature, material parameters, and water cooling parameters as the input of the initial artificial neural network model and the dam temperature field results as the output, fine-tune the neural network weights to achieve the transfer of the surrogate model from the prior bay to the actual pouring bay.

[0041] This embodiment can quickly construct a surrogate model through transfer learning. The prior bay can be constructed before obtaining the concrete dam model. The prior bay with lower calculation cost is used to provide a prior data set, and then a small amount of dam data set is supplemented. The prior bay surrogate model is transferred to the actual working conditions, so as to quickly obtain the calculation results of numerical simulation, solve the problem that the traditional surrogate model training process is complex and time-consuming, effectively reduce the difficulty of the optimization problem, and meet the requirements of the tight construction period in actual projects.

[0042] S4. Dual-objective optimization of water cooling: Construct a dual-objective optimization mathematical model for water cooling with the maximum internal temperature and water cooling cost, and use the NSGA-II algorithm to solve the temperature field surrogate model of the actual pouring bay of the concrete dam for each bay to obtain the Pareto set. The solution with the minimum distance from the utopia point is selected as the relative optimal solution through the equilibrium point method, so as to obtain the relative optimal water cooling scheme for the pouring bay.

[0043] Specifically, the dual-objective optimization mathematical model for water cooling uses the maximum temperature in the pouring bay as the measurement standard for safety performance, and the total cooling cost as the evaluation index for economic benefits. To prevent cracks in the concrete due to sudden temperature drop inside, the cooling rate and temperature gradient are used as constraints. The dual-objective optimization mathematical expression is:

[0044] In the formula: f ( x ) is the optimization objective, g ( x ) is the optimization constraint, T max is the maximum temperature of the pouring bay; C is the water cooling cost; T ( t n ) is t nTemperature at a moment; T w is the initial water temperature; usually the river water temperature; T o is the cooling water temperature; Q is the total cooling water flow rate, T v is the cooling rate; T w is the initial water temperature; T i is the temperature at the current moment; T i+1 is the temperature at the current moment; T min is the minimum temperature of the pouring bin; L is the horizontal distance between the maximum temperature and the minimum temperature; δ is the temperature gradient threshold.

[0045] Specifically, the Pareto set is a set of solutions in multi-objective optimization that cannot improve one objective without deteriorating other objectives. The Pareto set has non-dominance, that is, any solution in the set is not simultaneously outperformed by other solutions in all objectives.

[0046] Specifically, the equilibrium point method first defines the point where both objectives reach the minimum as the utopia point (ideal solution). After normalizing each optimization objective, calculate the distance between each point on the Pareto front and the utopia point, and select the solution with the closest distance as the relatively optimal solution. The expression is as follows:

[0047] In the formula: x i , y i are the horizontal and vertical coordinates of each scatter point in the Pareto set; x Utopia and y Utopia are the horizontal and vertical coordinates of the utopia point.

[0048] In this embodiment, an optimization control objective function is established based on safety and economy indicators, a double-objective optimization model is constructed, and the NSGA-II algorithm is used for intelligent optimization. The optimization results are as Figure 4 shown, solving the problem that the traditional method fails to fully consider the balance between economy and safety, improving the overall efficiency of the project, and providing a new solution for the temperature control technology of concrete dams.

[0049] S5. Model dynamic update: As the pouring process progresses, the temperature monitoring data gradually increases. The parameter identification technology is used to dynamically update the thermal parameters of the model to continuously improve the simulation accuracy, thereby ensuring the effectiveness of the optimization results.

[0050] Specifically, the parameter identification technology dynamically inverses the thermal parameters of concrete materials based on the monitoring data generated during the pouring process, where the thermal parameters are selected as one or several combinations of the final adiabatic temperature rise θ0, the number of days for half of the hydration heat n, and the surface heat release coefficient β.

[0051] S6. Repeat steps S3 - S5 until the optimization of all concrete dam pouring bins is completed.

[0052] In this embodiment, through the dynamic model update technology, the thermal parameters are calibrated in real time during the optimization process. The results after calibration are as Figure 5 shown, effectively reducing the deviation between the numerical calculation model and the actual temperature field, achieving an accurate match between the surrogate model and the actual working conditions, and ensuring the accuracy and reliability of the results.

[0053] The above specific implementation manners are the preferred implementation manners of the present invention, and do not limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation manner. All equivalent changes made according to the shape, structure, and method of the present invention are within the protection scope of the present invention.

[0054] As a preferred embodiment, as Figure 2 、 Figure 3 shown, it is a detailed explanation (principles, etc.) of the effects, purposes, or certain features of the above embodiments of the present invention and an expansion of the content of the implementation manners. Taking the first-stage water filling of the roller compacted concrete gravity dam of Huangdeng Hydropower Station as an example, a water cooling optimization control method for concrete dams based on transfer learning is proposed, including the following steps: Step 1: Select a reasonable range of prior bin dimensions, pouring temperature, air temperature, material parameters, and water cooling parameter variation ranges according to the actual dam construction situation. The reasonable ranges of the dam body pouring temperature, air temperature, material parameters, and water cooling parameter variation ranges are shown in Table 1. A sample design space is constructed through Latin hypercube sampling. A total of 2000 groups of samples are sampled for the prior bin, and 200 groups of training samples are sampled for the dam body.

[0055] Table 1 Parameter Variation Range Table

[0056] Step 2: With the help of ANSYS finite element analysis software, a simulation model of the prior bin is established by parametric modeling. Input relevant thermodynamic parameters, set boundary conditions, and perform temperature field calculation and analysis. The prior bin sample space points and the temperature field calculation results form the prior bin training samples. An artificial neural network surrogate model is used to establish the non-linear relationship between the prior bin boundary conditions and the temperature field, and a prior bin surrogate model is constructed.

[0057] Step 3: With the help of ANSYS finite element analysis software, based on the finite element model of the dam body, input relevant thermodynamic parameters, set boundary conditions, and perform temperature field calculation and analysis. The dam body sample space points and the temperature field calculation results are used to form the dam body training samples. Freeze some weights of the prior bin surrogate model, supplement the dam body training samples, and retrain the surrogate model to obtain the surrogate model of the actual pouring bin temperature field of the concrete dam.

[0058] Step 4: With the maximum temperature inside the dam body pouring bin and the water injection cost, and controlling the cooling rate and temperature gradient as the constraint conditions, construct a double-objective optimization mathematical model for water injection cooling. Use the surrogate model instead of the finite element simulation for the evaluation of the NSGA-II algorithm. The NSGA-II parameters are set as follows: the population size is 400, the maximum number of iterations is 400, the simulated binary crossover (SBX) is 0.9, and the polynomial mutation (PM) is 0.15. After normalizing the maximum temperature and water injection cost of the obtained population, select the point closest to the coordinate (0, 0) as the optimal water injection cooling measure.

[0059] Step 5: As the pouring process progresses, the temperature monitoring data gradually increases. According to the actual project, parameter identification is carried out every about 20 days to improve the accuracy of finite element calculation. The temperature monitoring data from May 23 to June 16, 2015 is used for the first model update. The data from June 16 to July 7, 2015 is supplemented for the second model update. And so on. Step 6: Repeat Steps 3 - 5 until all the pouring bins of the concrete dam are optimized. Subsequently, substitute the optimized water injection cooling scheme into the finite element calculation to obtain the Figure 6 temperature distribution results as shown, which verifies the effectiveness of this method. In addition, Figure 7 it further proves the advantage of this method in terms of calculation efficiency.

[0060] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with a processor function. Therefore, in engineering practice, the technical solutions and their functions of each embodiment of the present invention are encapsulated into various modules. Based on this actual situation, on the basis of the above embodiments, an embodiment of the present invention provides a transfer learning-based optimized control system for water injection cooling of a concrete dam, which is used to execute a transfer learning-based optimized control method for water injection cooling of a concrete dam in the above method embodiment.

[0061] The system includes: a first calculation module, which is used to select the prior bin size range, pouring temperature, air temperature, material parameters, and the change range of water-cooling parameters according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling; a second calculation module, which is used to establish a simulation model of the prior bin by using parametric modeling, obtain the calculation results of the temperature field through simulation, and use an artificial neural network surrogate model to establish the non-linear relationship between the boundary conditions of the prior bin and the temperature field, and construct a surrogate model of the temperature field of the prior bin; a third calculation module, which is used to use the simulation calculation data of the temperature field of the actual pouring bin of the concrete dam as the training sample, correct the surrogate model of the temperature field of the prior bin by using the simulation calculation results of the actual temperature field of the dam, and transfer and learn the surrogate model of the temperature field of the prior bin to the actual pouring bin of the dam to obtain a surrogate model of the temperature field of the actual pouring bin of the concrete dam; a fourth calculation module, which is used to construct a double-objective optimization mathematical model of water-cooling with the internal maximum temperature and water-cooling cost, use the NSGA-II algorithm to solve the surrogate model of the temperature field of the actual pouring bin of the concrete dam for each bin to obtain the Pareto set, and select the solution with the minimum distance from the utopia point as the relative optimal solution through the equilibrium point method to obtain the relative optimal water-cooling scheme for the pouring bin; a fifth calculation module, which is used to dynamically update the thermal parameters of the model by using parameter identification technology as the temperature monitoring data gradually increases during the pouring process.

[0062] An optimized control system for water-cooling of a concrete dam with transfer learning provided by an embodiment of the present invention aims at the problem that it is difficult to quickly and accurately optimize the existing water-cooling control method for concrete dams. By using the foregoing several modules, based on the simulation calculation model, a sample set of the temperature field of the prior bin is constructed. The non-linear relationship between the input parameters and the calculation results of the temperature field of the prior bin is captured through an artificial neural network, a surrogate model of the prior bin is established, and it is transferred and applied to the actual dam bin to significantly reduce the number of samples required for simulation calculation. Subsequently, with the maximum temperature and cooling cost as the optimization objectives, combined with the cooling rate and temperature gradient as the constraint conditions, a double-objective optimization model for water-cooling is constructed, and the NSGA-II algorithm is used to intelligently optimize and select the water-cooling scheme for the dam bin. During the optimization process, the thermal parameters are calibrated through the dynamic model update technology, further improving the reliability and accuracy of the optimization results. This system can generate accurate water-cooling measures that take into account both economy and safety in a short time, effectively solving the contradiction between tight construction period and insufficient optimization efficiency in actual projects, and has wide engineering application value.

[0063] It should be noted that the system embodiments provided by the present invention are used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The difference lies only in setting corresponding functional modules. The principle is basically the same as that of the above system embodiments provided by the present invention. As long as those skilled in the art, on the basis of the above system embodiments, refer to the specific technical solutions in other method embodiments, obtain corresponding technical means by combining technical features, and the technical solutions constituted by these technical means, and on the premise of ensuring the practicability of the technical solutions, improve the modules in the above system embodiments to obtain corresponding system-like embodiments for implementing the methods in other method-like embodiments.

[0064] Based on the same inventive concept as the foregoing embodiments, the embodiments of the present invention further provide an optimized control device for water cooling of a concrete dam in transfer learning, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the optimized control method for water cooling of a concrete dam in transfer learning.

[0065] Based on the same inventive concept as the foregoing embodiments, the embodiments of the present invention further provide a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the optimized control method for water cooling of a concrete dam in transfer learning.

[0066] In summary of the above embodiments, the present invention aims at the problem that the existing control methods for water cooling of concrete dams are difficult to optimize quickly and accurately, and proposes an optimized control method for water cooling of concrete dams based on transfer learning. This method is based on a simulation calculation model, constructs a prior bin temperature field sample set, captures the non-linear relationship between input parameters and the calculation results of the prior bin temperature field through an artificial neural network, establishes a prior bin surrogate model, and migrates and applies it to the actual dam bin to significantly reduce the number of samples required for simulation calculation. Subsequently, based on the maximum temperature and cooling cost as the optimization objectives, combined with the cooling rate and temperature gradient as the constraint conditions, a two-objective optimization model for water cooling is constructed, and the NSGA-II algorithm is used to intelligently optimize and select the water cooling scheme for the dam bin. During the optimization process, the thermal parameters are calibrated through a dynamic model update technology, further improving the reliability and accuracy of the optimization results. The present invention can generate precise water cooling measures that take into account both economy and safety in a short time, effectively solving the contradiction between tight construction period and insufficient optimization efficiency in actual projects, and having broad engineering application value.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the water cooling control of a concrete dam by transfer learning, characterized in that Including: S1. Select the prior bin size range, pouring temperature, air temperature, material parameters, and the change range of water-cooling parameters according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling; S2. Adopt parametric modeling to establish a simulation model of the prior bin, obtain the calculation results of the temperature field through simulation, and use the artificial neural network surrogate model to establish the non-linear relationship between the boundary conditions of the prior bin and the temperature field, and construct the prior bin temperature field surrogate model; S3. Use the simulation calculation data of the temperature field of the actual pouring bin of the concrete dam as the training sample, correct the prior bin temperature field surrogate model by using the simulation calculation results of the actual temperature field of the dam, and transfer the prior bin temperature field surrogate model to the actual pouring bin of the dam to obtain the concrete dam actual pouring bin temperature field surrogate model; S4. Construct a water-cooling double-objective optimization mathematical model with the maximum internal temperature and water-cooling cost, use the NSGA-II algorithm to solve the concrete dam actual pouring bin temperature field surrogate model for each bin to obtain the Pareto set, and select the solution with the minimum distance from the utopia point as the relative optimal solution through the equilibrium point method to obtain the relative optimal water-cooling scheme for the pouring bin; S5. As the pouring process progresses, the temperature monitoring data gradually increases, and the parameter identification technology is used to dynamically update the thermal parameters of the model; S6. Repeat steps S3 - S5 until all the pouring bins of the concrete dam are optimized.

2. The optimized control method for water cooling of concrete dams using transfer learning according to claim 1, characterized in that The said S2 further includes: S21. Obtain the geometric parameters of the dam construction; S22. Establish a typical rectangular dam bin according to the geometric parameters, and extend the dam bin downward by 1.5 times the dam height as the dam foundation; S23. Obtain the air temperature, material parameters, and water-cooling parameters as the input for the prior bin temperature field simulation calculation to obtain the temperature field calculation results; S24. Use the geometric parameters, air temperature, material parameters, and water-cooling parameters as the input of the artificial neural network model, and the temperature field results as the output to construct the initial artificial neural network surrogate model.

3. The optimized control method for water cooling of concrete dams using transfer learning according to claim 1, characterized in that, The said S23 further includes: Create a prior bin by parametric modeling or create a three-dimensional calculation model according to the project drawings; Determine the boundary conditions, thermodynamic parameters, and initial conditions according to the sample space generated by the concrete, and conduct temperature field calculation; Integrate the output temperature field simulation results, and integrate the sample space points and the output temperature field to form a training sample set.

4. The optimized control method for water cooling of concrete dams using transfer learning according to claim 1, characterized in that, The said S3 further includes: S31. Obtain the air temperature, material parameters, and water-cooling parameters as the input for the actual pouring bin temperature field simulation calculation to obtain the temperature field calculation results; S32. Use the air temperature, material parameters, and water-cooling parameters as the input of the initial artificial neural network model, and the dam body temperature field results as the output, and fine-tune the neural network weights to realize the transfer of the surrogate model from the prior bin to the actual pouring bin.

5. The optimized control method for the water cooling of a concrete dam using transfer learning according to claim 1, characterized in that, The said S4 further includes: The water-cooling double-objective optimization mathematical model uses the maximum temperature in the pouring bin as the measurement standard for safety performance, takes the total cooling cost as the evaluation index for economic benefits, and controls the cooling rate and temperature gradient as the constraint conditions. The double-objective optimization mathematical expression is: ， In the formula: f ( x ) is the optimization objective, g ( x ) is the optimization constraint, T max is the maximum temperature of the pouring bin; C is the water passing cost; T ( t n ) is t n the temperature at time T w is the initial water temperature; T o is the cooling water temperature; Q is the total cooling water flow rate, T v is the cooling rate; T i is the temperature at the current time; T i+1 is the temperature at time i + 1; T min is the minimum temperature of the pouring bin; L is the horizontal distance between the maximum temperature and the minimum temperature; δ is the temperature gradient threshold; is the temperature gradient.

6. The optimized control method for the water cooling of a concrete dam using transfer learning according to claim 5, characterized in that, Select the solution with the minimum distance from the utopia point as the relative optimal solution through the equilibrium point method, and further includes: Define the point where both objectives reach the minimum as the utopia point. After normalizing each optimization objective, calculate the distance between each point on the Pareto front and the utopia point, and select the solution with the closest distance as the relative optimal solution. The expression is as follows: ， Wherein: x i , y i are the abscissa and ordinate of each scatter point in the Pareto set; x Utopia and y Utopia are the abscissa and ordinate of the utopia point.

7. The optimized control method for water cooling of concrete dams using transfer learning according to claim 5, characterized in that, Adopt parameter identification technology to dynamically update the thermal parameters of the model, and it also includes: Dynamically invert the thermal parameters of concrete materials based on the monitoring data generated during the pouring process, where the thermal parameters are selected from the final adiabatic temperature rise θ 0, the number of days when the heat of hydration reaches half n , the surface heat release coefficient β or a combination of several of them.

8. An optimized control system for the water cooling of a concrete dam using transfer learning, characterized in that, including: The first calculation module is used to select the prior bin size range, pouring temperature, air temperature, material parameters, and the change range of water cooling parameters according to the actual dam construction situation, and construct a sample design space through Latin hypercube sampling; The second calculation module is used to establish a simulation model of the prior bin by using parametric modeling, obtain the temperature field calculation results through simulation, and use an artificial neural network surrogate model to establish the non-linear relationship between the boundary conditions of the prior bin and the temperature field, and construct a temperature field surrogate model for the prior bin; The third calculation module is used to use the temperature field simulation calculation data of the actual concrete dam pouring bin as the training sample, correct the prior bin temperature field surrogate model by using the actual temperature field simulation calculation results of the dam, and transfer the prior bin temperature field surrogate model to the actual concrete dam pouring bin to obtain the temperature field surrogate model of the actual concrete dam pouring bin; The fourth calculation module is used to construct a double-objective optimization mathematical model for water cooling with the internal maximum temperature and water cooling cost, use the NSGA-II algorithm to solve the temperature field surrogate model of the actual concrete dam pouring bin for each bin to obtain the Pareto set, and select the solution with the minimum distance from the utopia point as the relative optimal solution by the equilibrium point method to obtain the relative optimal water cooling scheme for the pouring bin; The fifth calculation module is used to dynamically update the thermal parameters of the model by using parameter identification technology as the temperature monitoring data gradually increases during the pouring process.

9. An optimized control device for the water cooling of a concrete dam using transfer learning, characterized in that, It includes a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of a method for optimizing the water cooling control of a concrete dam by transfer learning according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of a method for optimizing the water cooling control of a concrete dam by transfer learning according to any one of claims 1 to 7.

Images