A bivariate B-spline function surface solar shape model based on observational data

By using a bivariate B-spline function model of the Earth's surface solar shape based on observational data, the problems of complexity and fitting accuracy of existing models are solved, and high-precision simulation and efficient calculation of the radiation energy density distribution in solar thermal power generation systems are achieved.

CN114662403BActive Publication Date: 2024-11-01ZHEJIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210364209.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-07
Publication Date
2024-11-01
Estimated Expiration
2042-04-07

AI Technical Summary

Technical Problem

Existing models of the Earth's surface solar shape are complex, lack analytical integration of radiant energy, have low fitting accuracy, fail to align CSRs, and exhibit model discontinuities. These issues affect the accuracy and efficiency of radiant energy density distribution simulations for solar thermal power generation systems.

Method used

A bivariate B-spline function model of the Earth's surface solar shape based on observational data is adopted. Through measurement data processing and fitting of univariate and bivariate B-spline functions, a concise analytical model is constructed to ensure CSR alignment and high fitting accuracy.

Benefits of technology

This improves the accuracy and speed of radiative energy density distribution simulation in solar thermal power generation systems, simplifies the model structure, and achieves analytical integration of radiative energy and model continuity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114662403B_ABST
    Figure CN114662403B_ABST
Patent Text Reader

Abstract

The present invention discloses a bivariate B-spline function surface solar shape model based on observation data. The present invention first processes the surface solar intensity distribution observation data to obtain 84 groups of solar scanning profiles; then uses a bicubic B-spline function to perform fitting in two steps: the first step is to fit the 84 groups of solar scanning profile data along the angle offset direction to obtain intermediate control vertices, and the second step is to fit the intermediate control vertices to obtain the final control vertices. The surface solar shape analytical model established by the present invention has the advantages of simple model, analytically integral radiation energy, high fitting accuracy, CSR alignment and model continuity. The model is simple, conforms to the real solar intensity distribution, and can improve the accuracy and speed of the simulation of the radiation energy density distribution of the photothermal solar power generation system. The invention has a good application value in the field of photothermal solar energy simulation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of solar thermal power generation, and in particular to a surface sun shape model established based on real solar intensity distribution observation data. Background Art

[0002] In the field of solar thermal power generation system simulation, the sun shape model that describes the distribution of solar light intensity from the center to the edge is related to the accuracy of the simulation of the radiation energy density distribution on the receiver. Constructing an accurate sun shape model is conducive to obtaining accurate simulation results of the radiation energy density distribution. This result is not only the basis for optimizing the layout of the heliostat field in the early design stage of the tower solar thermal power generation system, but also an important guarantee for optimizing the focusing of the heliostat field, improving energy utilization, and protecting the receiver from damage in the later operation stage. The existing sun shape models are divided into two categories: extraterrestrial sun shape models and surface sun shape models. The extraterrestrial sun shape model does not take into account the intensity distribution of the solar halo, which will lead to an overestimation of the peak and total power in the simulation. At present, some existing surface solar shape models are simplified models, such as the Gaussian model, which deviates far from the actual solar intensity distribution. The surface solar shape model based on observational data is the model that best fits the actual solar intensity distribution. However, this type of model still has many problems, such as the Buie model and the B-spline splicing model have problems with insufficient fitting accuracy, model discontinuity, and CSR alignment (the CSR input to the model is not equal to the CSR calculated according to the model). Although the univariate B-spline function model solves the above problems, the model is complex and requires the storage of a large number of B-spline function control vertices. In particular, the Buie model also has the problem that the radiation energy calculation cannot be analytically integrated.

[0003] Taking into account the shortcomings of existing surface sun shape models, the present invention studies a surface sun shape modeling method based on real solar intensity observation data. Summary of the invention

[0004] The purpose of the present invention is to propose a bivariate B-spline function surface solar shape model based on observation data to address the problems of complex surface solar shape model, inability to analytically integrate radiant energy, low model fitting accuracy, inability to align CSR and model discontinuity.

[0005] The object of the present invention is achieved through the following technical scheme: A bivariate B-spline function surface sun shape model based on observation data is constructed by the following method:

[0006] Step (1), measurement data processing, using the surface discrete sun shape data measured by LBL laboratory and DLR laboratory, averaging and data splicing according to CSR grouping, and finally obtaining 84 groups of sun shape discrete scanning profiles corresponding to different CSRs;

[0007] Step (2), using a univariate cubic B-spline function to fit the 84 sets of data along the radial angle offset direction;

[0008] Step (3), based on the control vertices obtained by fitting in step (2), fitting is performed along the CSR direction.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: the surface sun shape analytical model established by the present invention based on the real surface sun intensity distribution observation data and using the bivariate cubic B-spline function has the advantages of simple model, analytical integration of radiant energy, high fitting accuracy, CSR alignment and model continuity. The model is simple and conforms to the real sun intensity distribution, and can improve the accuracy and speed of the simulation of the radiant energy density distribution of the solar thermal power generation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 This is the result diagram of DLR and LBL data processing;

[0011] Figure 2 It is the control vertex and control polygon graph corresponding to the two-step fitting;

[0012] Figure 3 It is the CSR alignment diagram of the bivariate B-spline function fitting of the differential evolution algorithm with different numbers of control vertices in the CSR direction, where (a) is the CSR alignment diagram when the number of control vertices is 9, (b) is the CSR alignment diagram when the number of control vertices is 10, (c) is the CSR alignment diagram when the number of control vertices is 11, and (d) is the CSR alignment diagram when the number of control vertices is 12;

[0013] Figure 4 is the result diagram of the bivariate B-spline function model, where (a) is the partial fitting result in the semi-logarithmic coordinate system, and (b) is the full fitting result in the spatial rectangular coordinate system;

[0014] Figure 5 are the model error comparison diagrams, where (a) is the maximum absolute value error diagram and (b) is the root mean square error diagram. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described below in conjunction with embodiments and accompanying drawings.

[0016] Example

[0017] Assume that the analytical model of the sun shape is:

[0018]

[0019] Where θ represents the angle offset, x represents the CSR, and Ni,d (θ) represents the B-spline basis function along the θ direction, N i,d (χ) represents the B-spline basis function along the x direction, P ij Represents the B-spline function control coefficient. Here, the unit of θ is mrad.

[0020] The total energy of the disc and halo parts can be derived as follows:

[0021]

[0022] Among them, Φ disk Here, the unit of θ is rad. When θ is sufficiently small, sin(θ)≈θ, cos(θ≈1, so

[0023]

[0024] Similarly,

[0025]

[0026] Among them, Φ aureole Represents the total radiant energy of the halo part, θ disk Indicates the maximum angular displacement of the central disk of the sun, which is 4.65 mrad, θ aureole Indicates the maximum angular offset of the peripheral halo, which is 43.6mrad.

[0027] CSR(χ) is defined as the total radiant energy Φ of the halo part aureole and total radiation energy Φ all The ratio of total radiation energy Φ all Refers to the radiant energy including the disc and halo.

[0028]

[0029] The bivariate B-spline function model construction in this paper includes the following steps:

[0030] S1, observation data processing. Currently, the authoritative and public data sets include the surface solar intensity distribution data observed by the Lawrence Berkeley Laboratory (LBL) and the German Aerospace Center (DLR). However, the LBL laboratory observation record data is not completely correct, and some processing operations are required, including the elimination of erroneous data, the elimination of data that deviates greatly from the average distribution, and the averaging based on CSR grouping. Finally, the processed LBL halo solar intensity distribution data is spliced ​​with the DLR CSR = 5% optical disc solar intensity distribution data. In the end, a total of 84 groups of solar intensity scanning profiles with CSR intervals of 0.01 from 0.005 to 0.835 are obtained. Each profile has 54 data points. The results are as follows Figure 1 Specifically, it is divided into the following steps:

[0031] Step (1), remove obviously wrong data. Discard data marked as 'Aperture in wrong position', 'Rain flap closed', 'Sun not centred in scan', and remove data with any negative value of SolRad, Circum, C / (C+S);

[0032] Step (2), remove the data with normalized light intensity far greater than 1. The normalization processing method is: each set of observation data is simultaneously divided by the light intensity of this set of data at the center position, so that the normalized data with the light intensity of the center of the sun as 1 is obtained. After normalization, there are some data whose light intensity at non-center positions is greater than or far greater than 1. If the maximum normalized sunlight intensity is greater than 1.05, then this set of data is removed and discarded;

[0033] Step (3), remove the data that deviates greatly from the average distribution. The data processed in steps (1) and (2) are averaged according to the CSR grouping. If the angle deviation is less than 3.7 mrad, if the observed data deviates from the mean by 10% above or below, the data is removed;

[0034] Step (4), CSR is divided into 100 equal parts from 0 to 1, and the data obtained by steps (1), (2) and (3) are first grouped according to the CSR size and then averaged. The averaging operation can reduce the measurement error caused by the noise of the measurement data;

[0035] Step (5), after group averaging, when CSR is greater than 0.835, some data have been completely eliminated, and some data have obvious fluctuations, which means that the original data has a measurement problem when CSR is greater than 0.835, or the light intensity is very random at this time. Therefore, the data with CSR greater than 0.835 is discarded. Finally, 84 groups of data are obtained, which are strictly monotonically decreasing with the increase of angle offset.

[0036] Step (6), data splicing. The first five steps are to process the LBL measurement data. This step is to splice the processed LBL halo solar intensity distribution observation data with the DLR CSR = 5% optical disk solar intensity distribution observation data based on the data obtained by LBL processing. After the above analysis and processing, a total of 84 sets of solar intensity scanning profiles with CSR ranging from 0.005 to 0.835 at intervals of 0.01 are finally obtained, and each profile has 54 data points. The results are as follows Figure 1 shown.

[0037] S2, for each of the 84 sets of data, a univariate cubic B-spline function was used to fit along the θ direction.

[0038] For any given χ k , k = 0, 1, ..., r,

[0039]

[0040] By fitting the B-spline curve along the θ direction, we can get R k,j , j = 0, 1, ..., n. Here, the univariate B-spline function model is used as the fitting result of the first step, such as Figure 2 (For ease of observation, the normalized sunlight intensity direction is logarithmic coordinates) as shown in the figure. The curve along the angle offset direction is given by k , the control polygons of B-spline curve fitting, a total of 84 sets of control polygons are obtained.

[0041] S3, based on the control vertices obtained by fitting in step (2), fitting is performed along the CSR direction respectively.

[0042] Satisfy P ij ≥0

[0043] Given the number and position of node vectors, use the above formula to calculate

[0044] R 0,j , R 1,j , ..., R r,j , j = 0, 1, ..., n

[0045] By fitting, we can get Pi,j ,i=0,1,...,m,j=0,1,...,n.

[0046] The position of the node vector in the CSR direction is determined using a differential algorithm, assuming that the node vector is:

[0047] v 0 , v 1 , ..., v d , v d+1 , ..., v m , v m+1 , ..., v m+d+1

[0048] An individual in the population is represented as:

[0049] V i,G =(υ i,G,0 ,…,υ i,G,d , i,G,d+1 , i,G,d+2 ,…,υ i,G,m+d+1 ), i = 0, 2, ..., NP-1

[0050] Experimental analysis shows that interpolating the left endpoint of the fitting result is beneficial to CSR alignment, that is, interpolation In addition, a large number of experimental results show that the right end point of the interpolation It is helpful to reduce the error. Therefore, the population individual node vector needs to satisfy:

[0051] υ i,G,0 =υ i,G,1 =...=υ i,G,d = x 0 =0.005

[0052] υ i,G,m+1 =υ i,G,m+2 =...=v i,G,m+d+1 = x r =0.835

[0053] From the monotonic property of the B-spline node vector, we know that each individual also needs to satisfy the following constraints:

[0054] 0.005 = v i,G,d ≤υ i,G,d+1 ≤υ i,G,d+2 ≤…≤υ i,G,m ≤υ i,G,m+1 =0.835

[0055] The uniform node vector generation method was used to generate a preliminary solution V pre Based on the preliminary solution, the first generation population is obtained by adding normal distribution random deviation

[0056] V i,1 (d+1:m)=U pre (d+1:m)+N(0,1),i=0,2,…,NP-1

[0057] V obtained after adding normal distribution random deviation i,1 The individual constraints may no longer be satisfied and further processing is required. First, v i,1,d+1 , v i,1,d+2 , …, v i,1,m For elements less than 0.05 or greater than 0.835, use [χ 0 , χ r ], and finally sort them from small to large. The result of the operation is still recorded as V under the premise that there is no misunderstanding i,1 In the mutation operation, similar individual constraint processing is also required.

[0058] Here, the population size NP = 40, the maximum evolutionary generation G m =2000, hybridization coefficient CR=0.4.

[0059] The termination condition is G<10.

[0060]

[0061] The goal of this paper is to select a result with the fewest number of control vertices, the most aligned CSRs, and the smallest fitting error. Figure 3 As shown, when the number of control vertices in the CSR direction is 9 and 10, the CSR is not aligned, and the CSR alignment effect when the number of control vertices is 11 is not as good as the alignment effect when the number of control vertices is 12.

[0062] like Figure 2 As shown in the figure, when the number of CSR direction control vertices is 12, the hollow circle in the figure represents the second step fitting control vertex P i,j , the curve along the CSR direction is the control polygon of the second step B-spline curve fitting.

[0063] Finally, the number of CSR direction control vertices is selected as 12, and the number of angle offset direction control vertices is selected as 15. Figure 4 are the fitting results, where the left figure is a partial fitting result in a semi-logarithmic coordinate system, and the right figure is the full fitting result in a spatial rectangular coordinate system. Figure 5 It is the fitting error diagram of different models. It can be found that the fitting errors of the bivariate B-spline function model and the univariate B-spline function model are basically the same.

Claims

1. A bivariate B-spline function surface sun shape model based on observation data, It is characterized in that It is constructed by the following method: Step (1), measurement data processing, using the surface discrete sun shape data measured by LBL laboratory and DLR laboratory, averaging and data splicing according to CSR grouping, and finally obtaining 84 groups of sun shape discrete scanning profiles corresponding to different CSRs; Step (2), using a univariate cubic B-spline function to fit the 84 sets of data along the radial angle offset direction; Step (3), based on the control vertices obtained by fitting in step (2), fitting is performed along the CSR direction respectively; The step (3) is specifically as follows: based on the control vertices obtained by fitting in step (2), fitting is performed along the CSR direction: Satisfy P ij ≥0 Among them, N i,d (χ) represents the B-spline basis function along the χ direction, P ij represents the control coefficient of the B-spline function; Given the number and position of node vectors, use the above formula to calculate R 0,j ,R 1,j ,...,R r,j,j =0,1,...,n By fitting, we can get P i,j ,i=0,1,...,m,j=0,1,...,n; The position of the node vector in the CSR direction is determined using a differential algorithm, assuming that the node vector is: v 0 ,v 1 ,…,v d ,v d+1 ,…,v m ,v m+1 ,…,v m+d+1 An individual in the population is represented as: V i,G =(v i,G,0 ,…,v i,G,d ,v i,G,d+1 ,v i,G,d+2 ,…,v i,G,m+d+1 ),i=0,2,…,NP-1 The population individual node vector must satisfy: v i,G,0 =v i,G,1 =…=v i,G,d =χ 0 =0.005 v i,G,m+1 =v i,G,m+2 =…=v i,G,m+d+1 =χ r =0.835 From the monotonic property of the B-spline node vector, we know that each individual also needs to satisfy the following constraints: 0.005=v i,G,d ≤in i,G,d+1 ≤in i,G,d+2 ≤…≤v i,G,m ≤in i,G,m+1 =0.835 The uniform node vector generation method was used to generate a preliminary solution V pre Based on the preliminary solution, the first generation population is obtained by adding normal distribution random deviation V i,1 (d+1:m)=V pre (d+1:m)+N(0,1),i=0,2,…,NP-1 V obtained after adding normal distribution random deviation i,1 The individual constraints may no longer be satisfied and further processing is required; first, v i,1,d+1 , v i,1,d+2 , …, v i,1,m For elements less than 0.05 or greater than 0.835, use [χ 0 , χ r ], and finally sort them from small to large. The result of the operation is still recorded as V under the premise that there is no misunderstanding. i,1 ; In the mutation operation, similar individual constraint processing is also required; Here, the population size NP = 40, the maximum evolutionary generation G m =2000, hybridization coefficient CR = 0.4; termination condition is G < 10; 2. The bivariate B-spline function surface sun shape model based on observation data according to claim 1, It is characterized in that The step (2) is specifically: For each of the 84 data sets, a univariate cubic B-spline function was used to fit along the θ direction; for any given χ k , k = 0, 1, ..., r: Perform B-spline curve fitting along the θ direction to obtain R k,j , j = 0, 1, ..., n, the univariate B-spline function model is used as the fitting result of the first step; Among them, φ(χ k , θ) represents the sun shape analytical model, θ represents the angle offset, χ represents CSR, N i,d (θ) represents the B-spline basis function along the θ direction, N i,d (x k ) represents the direction along χ k Directed B-spline basis functions.