Heliostat field multi-target thermal power output optimization method based on Monte Carlo algorithm and GP-GOMEA

Abstract

The invention discloses a multi-target thermal power output optimization method of a heliostat field based on a Monte Carlo algorithm and GP-GOMEA, and particularly relates to the field of heliostat field output power optimization, comprising the following steps: firstly, through step A1, establishing a proper coordinate system as an analysis basis; then, in the step A2, an accurate mathematical model of the heliostat field is constructed by using a GP-GOMEA algorithm; in order to predict and analyze the performance of the heliostat field, a Transform model is introduced in the step A3, and key data support is provided for optimization design; in order to further improve the optimization effect, the NSGA-II, the GP-GOMEA and the Transform are combined in the step A4 to form a multi-modal genetic model so as to solve the optimization design problem of the heliostat field; then, in the step A5, based on the evaluation function and the input data, an NSGA-II algorithm is used for solving, and a group of Pareto optimal solutions are obtained to serve as candidate schemes; finally, in the step A6, a final scheme is carefully selected from Pareto optimal solutions to serve as an optimization design result of the heliostat field.

Description

Technical Field

[0001] The present invention relates to the technical field of heliostat field output power optimization, and more specifically, to a heliostat field multi-objective thermal power output optimization method based on Monte Carlo algorithm and GP-GOMEA. Background Art

[0002] Heliostat field is an important part of solar thermal power generation system, and its optimization design is crucial to improve the thermal power output of the system. At present, Monte Carlo algorithm has been widely used in the optimization design of heliostat field. It estimates the optical efficiency and output thermal power of heliostat field through random sampling. In addition, nonlinear goal programming model is also used to maximize the output thermal power of heliostat per unit area.

[0003] However, there are still some shortcomings in its actual use. For example, although the Monte Carlo algorithm can provide effective optimization of the heliostat field, it usually requires a lot of computing resources and time to achieve high-precision results. In addition, traditional optimization methods may not fully consider the interactions between the heliostats in the heliostat field and complex environmental factors, resulting in less than ideal optimization results. Summary of the invention

[0004] In order to overcome the above defects of the prior art, an embodiment of the present invention provides a multi-objective thermal power output optimization method for a heliostat field based on a Monte Carlo algorithm and GP-GOMEA to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] Step A1: Establishment and selection of coordinate system;

[0007] Step A2: Use GP-GOMEA algorithm to construct a mathematical model of the heliostat field;

[0008] Step A3: Use the Transformer model to perform prediction and analysis to provide data support for the optimal design of the heliostat field;

[0009] Step A4: Combine NSGA-II, GP-GOMEA and Transformer to build a multimodal genetic model to solve the optimal design problem of the heliostat field;

[0010] Step A5: Using the NSGA-II algorithm to solve the optimization design problem of the heliostat field according to the evaluation function and input data, a set of Pareto optimal solutions are obtained as candidate solutions;

[0011] Step A6: Select a final solution from the Pareto optimal solutions as the optimization design result of the heliostat field.

[0012] Preferably, in step A1, the heliostat device is affected by the sun, the shadow of the absorption tower, and other heliostat devices when working. In order to better describe the working principle of the heliostat, a rectangular coordinate system is established with the center of the circular area of ​​the mirror field as the origin, the due east direction as the positive direction of the x-axis, the due north direction as the positive direction of the y-axis, and the vertical upward direction as the positive direction of the z-axis, which is called the mirror field coordinate system; then a rectangular coordinate system is established with the sun as the origin and the circular orbital plane of the earth as the xy plane, where the circular orbital plane of the earth is called the ecliptic, which is called the solar coordinate system, that is, the mirror rectangular coordinate system.

[0013] Preferably, in step A2, a mathematical expression describing the relationship between the thermal power output of the heliostat field and various parameters is automatically evolved as an evaluation function using the GP-GOMEA algorithm according to given data and constraints. The calculation method of the evaluation function is specifically as follows:

[0014] Where P represents the thermal power output of the heliostat field; DNI represents the direct solar radiation; SM represents the solar multiple; η helio Expressed as the heliostat field efficiency; η r Expressed as the efficiency of other power plant systems; H helio W is the length of the heliostat mirror. helio It is represented by the width of the heliostat mirror; ALT is the altitude; θ L It is represented by the angle between the line connecting the heliostat and the heat absorber and its projection on the horizontal plane; d is represented by the distance between the heliostat and the heat absorber;

[0015] The evaluation function is constructed into a mathematical model of the heliostat field.

[0016] Preferably, in step A3, a python crawler is used to obtain the daily solar normal direct radiation DNI from the weather network as input data for the heliostat field optimization design; the model can generate accurate and stable prediction results based on the seasonality, periodicity and trend of the data, as well as the correlation and causality between the data;

[0017] In order to use the Transformer model to predict the solar normal direct radiation DNI, historical data and weather forecasts are used as input sequences, and the DNI for each day in the next year is used as the output sequence.

[0018] Preferably, in step A4, the NSGA-II algorithm is used to solve the optimization design problem of the heliostat field according to the evaluation function and the input data, and a set of Pareto optimal solutions are obtained as candidate solutions. The algorithm can perform fast non-dominated sorting and crowding degree allocation on individuals according to the values ​​of multiple objective functions, thereby maintaining the diversity and elitism of the population.

[0019] Preferably, in step A5, the population size, crossover probability, mutation probability, maximum number of iterations, and method for initializing the population are randomly generated or use heuristic rules; the values ​​of these parameters will affect the convergence speed of the algorithm and the quality of the solution, and therefore need to be adjusted according to the characteristics of the problem and experience; generally speaking, the population size should be large enough to ensure the diversity and coverage of the population; the crossover probability and mutation probability should be moderate to balance the exploratory and developmental nature of the algorithm; the maximum number of iterations should be set according to the complexity of the problem and computing resources; the method for initializing the population can be randomly generated or use heuristic rules.

[0020] Preferably, in step A6, the calculation method in the selection process is specifically:

[0021] Among them, x * is the final solution; P is the Pareto optimal solution set; w1 is the priority of different goals, w2 is the preference of different goals; two weight coefficients; this formula indicates the selection of a trade-off solution that comprehensively considers multiple indicators, that is, to find a balance point between the annual average output thermal power per unit mirror area and the mirror field area;

[0022] Focusing on a certain target, the annual average output thermal power per unit mirror area, the corresponding weight coefficient is set to 1, and the others are set to 0. The specific calculation method is:

[0023]

[0024] In this way, we can get the solution with the maximum average annual thermal output per unit mirror area. Focus on the mirror area, set the corresponding weight coefficient to 1, and set the others to 0. The specific calculation method is:

[0025]

[0026] Finally, the solution with the smallest mirror field area can be obtained.

[0027] Technical effects and advantages of the present invention:

[0028] 1. By combining NSGA-II, GP-GOMEA and Transformer models, this patent proposes a multimodal genetic model that can effectively solve the optimization design problem of heliostat fields. This algorithm can not only handle complex multi-objective optimization problems, but also maintain the diversity and elitism of the population. The Pareto optimal solution obtained is evenly distributed and has good convergence and robustness;

[0029] 2. The GP-GOMEA algorithm can automatically discover and evolve mathematical expressions without manually setting the model structure and parameters. This enables the mathematical model of the heliostat field to adapt to the design of different sites and conditions, improving the versatility and adaptability of the model.

[0030] 3. The Transformer model uses the self-attention mechanism to capture the long-distance dependencies between data, and can accurately predict the solar radiation and solar position time series data, providing reliable data support for the optimal design of the heliostat field;

[0031] 4. This patent allows users to choose the final solution based on different priorities and preferences. Whether it is to maximize the annual average output thermal power per unit mirror area, minimize the mirror field area, or a trade-off between the two, the optimal solution that meets the needs can be found. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0033] Figure 2 It is a schematic diagram of the mirror field coordinate system of the present invention.

[0034] Figure 3 It is a schematic diagram of the position of the heliostat of the present invention.

[0035] Figure 4 It is a schematic diagram of light shielding of the absorption tower of the present invention.

[0036] Figure 5 Schematic diagram of the relative position relationship between the collector and the heliostat of the present invention DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] See also Figure 1 As shown, the present invention is a multi-objective thermal power output optimization method for a heliostat field based on a Monte Carlo algorithm and GP-GOMEA, comprising the following steps:

[0039] Step A1: Establishment and selection of coordinate system;

[0040] In step A1, the heliostat device is affected by the sun, the shadow of the absorption tower, and other heliostat devices when working. In order to better describe the working principle of the heliostat, a rectangular coordinate system is established with the center of the circular area of ​​the mirror field as the origin, the due east direction as the positive direction of the x-axis, the due north direction as the positive direction of the y-axis, and the vertical upward direction as the positive direction of the z-axis, which is called the mirror field coordinate system; then a rectangular coordinate system is established with the sun as the origin and the circular orbital plane of the earth as the xy plane, where the circular orbital plane of the earth is called the ecliptic, which is called the solar coordinate system, that is, the mirror rectangular coordinate system.

[0041] Step A2: Use GP-GOMEA algorithm to construct a mathematical model of the heliostat field;

[0042] In step A2, the GP-GOMEA algorithm is used to automatically evolve a mathematical expression describing the relationship between the thermal power output of the heliostat field and various parameters as an evaluation function according to the given data and constraints. The calculation method of the evaluation function is specifically as follows:

[0043] Where P represents the thermal power output of the heliostat field; DNI represents the direct solar radiation; SM represents the solar multiple; η helio Expressed as the heliostat field efficiency; η r Expressed as the efficiency of other power plant systems; H helio W is the length of the heliostat mirror. helio It is represented by the width of the heliostat mirror; ALT is the altitude; θ L It is represented by the angle between the line connecting the heliostat and the heat absorber and its projection on the horizontal plane; d is represented by the distance between the heliostat and the heat absorber;

[0044] The evaluation function is constructed into a mathematical model of the heliostat field.

[0045] Step A3: Use the Transformer model to perform prediction and analysis to provide data support for the optimal design of the heliostat field;

[0046] In step A3, a python crawler is used to obtain the daily solar normal direct radiation DNI from the weather network as input data for the optimization design of the heliostat field; the model can generate accurate and stable prediction results based on the seasonality, periodicity and trend of the data, as well as the correlation and causality between the data;

[0047] In order to use the Transformer model to predict the solar normal direct radiation DNI, historical data and weather forecasts are used as input sequences, and the DNI of each day in the next year is used as the output sequence. The specific method of building the model is as follows:

[0048] Data preprocessing: convert historical data and weather forecasts into numerical vectors using normalization, standardization, and discretization; at the same time, appropriately scale the DNI values ​​of the output sequence to facilitate model training and prediction;

[0049] Model training: Define the parameters of the Transformer model, number of layers, number of hidden units, number of attention heads, feedforward network dimensions, activation function. Use optimization algorithms, Adam 3 , and the loss function, mean square error MSE, to update the model weights; use some techniques to improve the performance of the model, learning rate decay, gradient clipping, regularization, dropout;

[0050] Model testing: Use the trained model to predict the test set or validation set, and use some evaluation indicators to measure the effect of the model;

[0051] Model application: Deploy the trained model to actual scenarios, generate DNI forecasts for each day of the next year based on real-time or recent historical data and weather forecasts, and use them as input data for heliostat field optimization design;

[0052] The calculation method for predicting DNI using the Transformer model is as follows:

[0053] B=(b1,b2,…,b T ), where B represents the input sequence and T represents the length of the input sequence;

[0054] Y=(y1,y2,…,y T′ ), where Y represents the output sequence and T′ is the length of the output sequence;

[0055] Z=(z1,z2,…,z T )=Encoder(X), where Z represents the encoder representation value, T represents the length of the input sequence, and Encoder is the encoder of the Transformer model;

[0056] in, Represents the encoder output value, T′ is the length of the output sequence, and Decoder represents the decoder of the Transformer model;

[0057] Where L is the loss function, θ is the weight of the model, T′ is the length of the output sequence, and y t is the DNI value at the tth time point, is the predicted value of DNI at the tth time point;

[0058] Among them, θ(k+1) It is represented as the optimization algorithm, θ is the weight of the model, α is the learning rate, It is represented as the gradient operator, and k is the number of iterations;

[0059] Transformer is a deep neural network model consisting of an encoder and a decoder, which can use the self-attention mechanism to capture long-range dependencies between data, thereby improving prediction accuracy and interpretability.

[0060] Step A4: Combine NSGA-II, GP-GOMEA and Transformer to build a multimodal genetic model to solve the optimal design problem of the heliostat field;

[0061] In step A4, the NSGA-II algorithm is used to solve the optimization design problem of the heliostat field according to the evaluation function and the input data, and a set of Pareto optimal solutions are obtained as candidate solutions. The algorithm can perform fast non-dominated sorting and crowding degree allocation on individuals according to the values ​​of multiple objective functions, thereby maintaining the diversity and elitism of the population.

[0062] The various parameters of the heliostat field, including the location coordinates of the absorption tower x1, the size of the heliostat x2, the installation height x3, the number of heliostats x4, and the location of the heliostat x5, are used as optimization variables. The annual average output thermal power per unit mirror area f(x) is used as the optimization objective function. The rated power P r And other constraints g(x)≤0 as optimization constraints, forming a calculation method for a multi-objective optimization problem:

[0063]

[0064] Among them, x is represented as a vector containing all optimization variables, f(x) is a function calculated based on optical efficiency and meteorological data, g(x) is a function defined according to physical laws and engineering requirements, P r is a given constant, representing the rated power of the heliostat field, and t0 is a time variable, representing the day of the year.

[0065] Step A5: Using the NSGA-II algorithm to solve the optimization design problem of the heliostat field according to the evaluation function and input data, a set of Pareto optimal solutions are obtained as candidate solutions;

[0066] In step A5, the population size, crossover probability, mutation probability, maximum number of iterations, and the method for initializing the population are randomly generated or use heuristic rules; the values ​​of these parameters will affect the convergence speed of the algorithm and the quality of the solution, and therefore need to be adjusted according to the characteristics of the problem and experience; generally speaking, the population size should be large enough to ensure the diversity and coverage of the population; the crossover probability and mutation probability should be moderate to balance the exploratory and developmental nature of the algorithm; the maximum number of iterations should be set according to the complexity of the problem and the computing resources; the method for initializing the population can be randomly generated or use heuristic rules;

[0067] Perform fast non-dominated sorting on each individual in the current population, that is, assign it to different non-dominated layers according to the comparison results of its objective function value with other individuals, where the first layer is the optimal layer, that is, no individual can be better than it in all objectives at the same time, the second layer is the suboptimal layer, that is, only the individuals in the first layer can be better than it in at least one objective, and so on; the specific method of fast non-dominated sorting is:

[0068] Set the current non-dominated layer number, set the number of dominant individuals and the set of dominated individuals of all individuals in the current population

[0069] Determine whether the termination condition is met; if the termination condition is met, stop the iteration and output all individuals in the optimal layer as the Pareto optimal solution set; a commonly used termination condition is to set a maximum number of iterations T max , that is, when the algorithm executes to the Tth max When the number of generations reaches , the iteration stops, and all individuals in the optimal layer of the current population are output as the Pareto optimal solution set.

[0070] Step A6: Select a final solution from the Pareto optimal solutions as the optimal design result of the heliostat field;

[0071] In step A6, the calculation method in the selection process is specifically:

[0072] Among them, x * is the final solution; P is the Pareto optimal solution set; w1 is the priority of different goals, w2 is the preference of different goals; two weight coefficients; this formula indicates the selection of a trade-off solution that comprehensively considers multiple indicators, that is, to find a balance point between the annual average output thermal power per unit mirror area and the mirror field area;

[0073] Focusing on a certain target, the annual average output thermal power per unit mirror area, the corresponding weight coefficient is set to 1, and the others are set to 0. The specific calculation method is:

[0074]

[0075] In this way, we can get the solution with the maximum average annual thermal output per unit mirror area. Focus on the mirror area, set the corresponding weight coefficient to 1, and set the others to 0. The specific calculation method is:

[0076]

[0077] Finally, the solution with the smallest mirror field area can be obtained.

[0078] Please read Figure 2 As shown, Figure 2 It is a schematic diagram of the mirror field coordinate system, an important diagram used in this patent to describe the position relationship of the heliostat field. In the patent document, the correct selection and establishment of the coordinate system is considered to be the key to accurately describe the working principle of the heliostat field. Specifically, the mirror field coordinate system diagram shows the position of each heliostat in the heliostat field and their positional relationship with the absorption tower (i.e., absorber).

[0079] See also Figure 3 As shown, Figure 3 It is a schematic diagram of the heliostat position, which intuitively shows the layout and relative position relationship of the heliostats in the mirror field. In the patent, the position and layout of the heliostats are key factors in the optimized design because they directly affect the optical efficiency and thermal power output of the mirror field.

[0080] The optimization methods proposed in the patent, including the mathematical model based on GP-GOMEA and the NSGA-II algorithm, are all centered around how to effectively determine the position and angle of the heliostats. The goal of these algorithms is to find the best heliostat configuration to maximize the thermal power output of the mirror field.

[0081] Figure 3 It provides a visual reference for the optimized design in the patent, helping to understand the working principle of the heliostat field and the spatial relationship that needs to be considered in the optimization process. It is closely related to the technical content in the patent and is an indispensable part of the patent application.

[0082] See also Figure 4 As shown, Figure 4 This is a schematic diagram of the absorption tower blocking light, which intuitively shows the blocking effect of the absorption tower (i.e., heat absorber) on the heliostat in the heliostat field.

[0083] See also Figure 5 As shown, Figure 5 This is a schematic diagram of the relative position relationship between the collector and the heliostat, which intuitively shows the relative position relationship between the collector (absorber) and the heliostat in the heliostat field. In the patent, this position relationship is a key factor to be considered in the optimization design because it directly affects the optical efficiency and thermal power output of the mirror field.

[0084] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-objective thermal power output optimization method for a heliostat field based on the Monte Carlo algorithm and GP-GOMEA, characterized in that: include: Step A1: Establishment and selection of coordinate system; Step A2: Use GP-GOMEA algorithm to construct a mathematical model of the heliostat field; Step A3: Use the Transformer model to perform prediction and analysis to provide data support for the optimal design of the heliostat field; Step A4: Combine NSGA-II, GP-GOMEA and Transformer to build a multimodal genetic model to solve the optimal design problem of the heliostat field; Step A5: Using the NSGA-II algorithm to solve the optimization design problem of the heliostat field according to the evaluation function and input data, a set of Pareto optimal solutions are obtained as candidate solutions; Step A6: Select a final solution from the Pareto optimal solutions as the optimization design result of the heliostat field.

2. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In the step A1, the heliostat device is affected by the sun, the shadow of the absorption tower, and other heliostat devices when working. In order to better describe the working principle of the heliostat, a rectangular coordinate system is established with the center of the circular area of ​​the mirror field as the origin, the due east direction as the positive direction of the x-axis, the due north direction as the positive direction of the y-axis, and the upward direction perpendicular to the ground as the positive direction of the z-axis, which is called the mirror field coordinate system; and then a rectangular coordinate system is established with the sun as the origin and the circular orbital plane of the earth as the xy plane, wherein the circular orbital plane of the earth is called the ecliptic, which is called the solar coordinate system, that is, the mirror rectangular coordinate system.

3. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In step A2, the GP-GOMEA algorithm is used to automatically evolve a mathematical expression describing the relationship between the thermal power output of the heliostat field and various parameters as an evaluation function according to given data and constraints.

4. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In step A3, a python crawler is used to obtain the daily solar normal direct radiation DNI from the weather network as input data for the optimization design of the heliostat field; the model can generate accurate and stable prediction results based on the seasonality, periodicity and trend of the data, as well as the correlation and causality between the data; In order to use the Transformer model to predict the solar normal direct radiation DNI, historical data and weather forecasts are used as input sequences, and the DNI of each day in the next year is used as the output sequence. The specific method of building the model is as follows: Data preprocessing: convert historical data and weather forecasts into numerical vectors using normalization, standardization, and discretization; at the same time, appropriately scale the DNI values ​​of the output sequence to facilitate model training and prediction; Model training: define the parameters of the Transformer model, number of layers, number of hidden units, number of attention heads, feedforward network dimensions, activation function; use optimization algorithm, Adam 3 , and the loss function, mean square error MSE, to update the model weights; Use some techniques to improve the performance of the model, such as learning rate decay, gradient clipping, regularization, and dropout; Model testing: Use the trained model to predict the test set or validation set, and use some evaluation indicators to measure the effect of the model; Model application: Deploy the trained model to actual scenarios, generate DNI forecasts for each day of the next year based on real-time or recent historical data and weather forecasts, and use them as input data for the optimal design of the heliostat field.

5. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In step A4, the NSGA-II algorithm is used to solve the optimization design problem of the heliostat field according to the evaluation function and the input data, and a set of Pareto optimal solutions are obtained as candidate solutions; The algorithm can perform fast non-dominated sorting and crowding degree allocation on individuals according to the values ​​of multiple objective functions, thereby maintaining the diversity and elitism of the population; The various parameters of the heliostat field, including the location coordinates of the absorption tower x1, the size of the heliostat x2, the installation height x3, the number of heliostats x4, and the location of the heliostat x5, are used as optimization variables. The annual average output thermal power per unit mirror area f(x) is used as the optimization objective function. The rated power P r and other constraints g(x)≤0 as optimization constraints.

6. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 5, characterized in that: The calculation method to form a multi-objective optimization problem is as follows: Among them, x is represented as a vector containing all optimization variables, f(x) is a function calculated based on optical efficiency and meteorological data, g(x) is a function defined according to physical laws and engineering requirements, P r is a given constant, representing the rated power of the heliostat field, and t0 is a time variable, representing the day of the year.

7. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In step A5, the population size, crossover probability, mutation probability, maximum number of iterations, and the method of initializing the population are randomly generated or use heuristic rules; the values ​​of these parameters will affect the convergence speed of the algorithm and the quality of the solution, so they need to be adjusted according to the characteristics of the problem and experience; generally speaking, the population size should be large enough to ensure the diversity and coverage of the population; The crossover probability and mutation probability should be moderate to balance the exploratory and exploitative nature of the algorithm; the maximum number of iterations should be set according to the complexity of the problem and computing resources; the method for initializing the population can be random generation or the use of heuristic rules.

8. The method for optimizing the multi-objective thermal power output of a heliostat field based on the Monte Carlo algorithm and GP-GOMEA according to claim 1, characterized in that: In step A6, the calculation method in the selection process is specifically: Among them, x * is the final solution; P is the Pareto optimal solution set; w1 is the priority of different goals, w2 is the preference of different goals; two weight coefficients; this formula indicates the selection of a trade-off solution that comprehensively considers multiple indicators, that is, to find a balance point between the annual average output thermal power per unit mirror area and the mirror field area; Focusing on a certain target, the annual average output thermal power per unit mirror area, the corresponding weight coefficient is set to 1, and the others are set to 0. The specific calculation method is: In this way, we can get the solution with the maximum average annual thermal output per unit mirror area; focus on the mirror area, set the corresponding weight coefficient to 1, and set the others to 0. The specific calculation method is: Finally, the solution with the smallest mirror field area can be obtained.

Images