Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group
Journal Home About Issues in Progress Current Issue All Issues Feature Issues

Design, fabrication, and evaluation of a large-area hybrid solar simulator for remote sensing applications

Zhiqiang Du, Huijie Zhao, Guorui Jia, and Xudong Li

Open Access Open Access

Abstract

Solar irradiance variations have a direct effect on the accuracy and repeatability of identifying spectral signatures in the remote sensing field experiments. Solar simulators have been deployed to allow for testing under controlled and reproducible laboratory conditions. However, it is difficult and expensive to make a large-area solar simulation with the appropriate spectral content and spatial uniformity of irradiance. In this study, a hybrid solar simulator has been designed and constructed to provide large-area illumination for remote sensing simulation applications. A design method based on the two-phase genetic algorithm is proposed to improve the performance of the spectral match and spatial uniformity, which no longer relies on the traditional trial-and-error technique. The first phase is used to determine the most appropriate configuration of different lamps in order to represent the solar spectrum. The second phase is to accommodate an optimal placement of the multiple sources to achieve irradiance uniformity. Both numerical simulations and experiments were performed to verify the performances. The results showed that the solar simulator provided a good spectral match and spatial irradiance for simulating the variations in direct normal irradiance at different solar zenith angles. In addition, the modular design makes it possible to adjust irradiance on the target area without altering the spectral distribution. This work demonstrates the development and measurement of a hybrid solar simulator with a realizable optimal configuration of multiple lamps, and offers the prospect of a scalable, large-area solar simulation.

© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement

1. Introduction

The sun provides the most common source of radiation for remote sensing applications [1]. The variations in solar irradiance directly can directly affect the performance of extracting and identifying spectral signatures of materials [2]. For example, illumination variations due to sensor-Sun geometry can bring uncertainties to the retrieval of phenological metrics, especially for three-dimensionally complex vegetation. The solar elevation angle has a significant effect on surface albedo variation characteristics and trends [3]. Visible and near-infrared albedos are more sensitive to solar zenith angles than azimuth angles [4]. Additionally, different regions of the solar spectrum offer many opportunities for monitoring biotic and abiotic plant stress in hyperspectral remote sensing studies. However, unstable weather conditions make it difficult to conduct repeatable field experiments based on natural sunlight [5]. In this context, solar simulators have been manufactured to provide spectral emission and spatial irradiance close to sunlight within the laboratory [6].

Thus far, solar simulators have been deployed in the fields of aerospace testing, photovoltaic devices calibration, and solar thermochemical research [7,8]. There are common artificial light sources used in solar simulators, such as various gas discharge lamps, halogen lamps, and light-emitting diodes (LEDs) [9,10]. Carbon arc light sources are used for multi-junction solar cell tests since the spectral structure is compatible with the sun's radiation in space [11]. Xenon arc lamps are the most frequently used in high-flux solar simulators, which provide a quasi-point- source coupled to a truncated ellipsoidal reflector, resulting in an axially symmetric radiative flux distribution at the focal plane. The approach to tailoring the radiative flux distribution involves realignment of lamps and reflectors, and application of secondary optics such as compound parabolic concentrators (CPCs) and beam shaping optical systems [9]. In addition, the emission spectrum of a xenon arc lamp resembles that of blackbody radiation at 6000 K and approximates the terrestrial solar spectrum, but there are many peaks in the spectrum [12]. High peaks in the Xe emission spectrum between 850 nm and 1000 nm create a significant discrepancy to the solar spectrum [8]. A lot of high-flux solar simulators based on single-source or multi-source have been developed and characterized for several years in the sector of solar thermal research [5, 13–15]. Metal halide lamps most closely follow the AM1.5D reference solar spectrum, and they are also widely used in high flux solar simulators because of their high luminous efficacy, relatively inexpensive price, and long operating lifetime [16]. Halogen lamps have been used to simulate the remote multispectral imaging process, but the emissive power distributions of a halogen lamp is quite different from the actual solar spectrum in the visible and ultraviolet bands [17]. While additional filters have been used to adjust the shape of the spectral distribution, this makes the system unnecessarily bulky and expensive [7]. To simulate the solar spectrum with good accuracy, different types of LEDs are usually assembled to match the solar spectrum in the 400 nm–1100 nm spectral range [18]. However, LED-based solar simulators are mainly used to measure small scale photovoltaics due to limitations of the low intensities, narrow emitting profiles and the availability of LED technology [19]. The research works involving multi-wavelength and multiple light source solar simulators have received extensive attention [20]. To overcome the shortcomings of a narrow spectra range and low intensity, a solar simulator was designed using LEDs and halogen lamps for the visible and infrared spectra [21]. Wang et al. proposed a numerical algorithm to obtain a theoretical estimation of the number of LED light sources with 14 wavelengths, and performed spectral fitting to meet the AM1.5 standard solar spectrum [22]. Stuckelberger et al. used a four-tile illumination module to improve the performance of a solar simulator with an 18 × 18 cm2 area [23]. A variety of solar simulators based on different light sources have been developed, but there are some specific requirements on spectral mismatch, spatial uniformity and illuminated area for the solar simulators used for remote sensing applications. Existing artificial light sources might be unsuitable for the assessment of vegetation in remote sensing applications due to the excessive heat radiated or narrow spectral range. The spectral mismatch and non-uniformity of illumination also lead to significant errors in hyperspectral imaging results. In addition, a large illuminated area and various incident angles are necessary for measuring reflectance in remote sensing physical simulation. There is a challenge to achieve the most appropriate solar simulation under specific conditions. The type, number, power, size, and emission characteristics of the light source, as well as the relative alignment of the source influence the radiative characteristics of a solar simulator [16]. Therefore, we will develop a method to determine the best solution for a solar simulator with a more accurate and uniform spectral irradiance output over large illumination areas.

In this paper, we report on the design and characterization of a hybrid solar simulator based on multiple sources. A two-phase genetic algorithm is proposed to optimize spectral match and irradiance uniformity. The algorithm provides an accurate selection of the types and numbers of lamps to meet spectrum matching criteria, and achieves an optimal arrangement of source array in order to meet spatial uniformity requirements. The complete optical set-up of the solar simulator has been described in detail and the performances have been analyzed. The main contribution of this study is to provide an automated methodology for choosing the appropriate solution for a solar simulation with good spectral match and uniformity. The constructed solar simulator provides a reliable and controlled environment for hyperspectral remote sensing studies and especially for the early pathogen detection in plants and spatio-temporal vegetation variations.

2. Method

2.1 Solar spectral irradiance characteristics

Remote sensing instruments are designed to detect energy over several separate wavelength ranges at various spectral resolutions. For example, the leaf pigmentations and mesophyll cell structures of vegetation are analyzed in the visible range (Vis, 350–700 nm) and near infrared range (NIR, 700–1100 nm). The water content of a plant can be detected in SWIR regions (SWIR, 1100–2500 nm) [24]. In this research, we need to simulate the solar spectrum over a wide range from 350 nm to 2500 nm. However, it is a difficult task to simulate the spectral distribution of sunlight. The solar spectral irradiance on the Earth’s surface varies by geographical location, time of day, climate conditions, and the state of the atmosphere [25]. The MODTRAN 4.1 radiative transfer code was used to analyze the direct solar irradiance. Three atmosphere profiles were used including the Mid-Latitude Summer (MLS), Sub-Arctic Summer (SAS), and Tropical (TRO) profiles [26]. Three aerosol types (rural, urban, and maritain aerosol) were chosen to simulate the characteristics of aerosol. The visibility condition was set as VIS = 25 km. The solar zenith angles were changed from 10° to 70° with an interval of 10°. In order to identify the typical spectral characteristics, the k-means clustering method was used to divide a given dataset into a cluster center and minimize the intracluster variance [27]. The direct solar spectral irradiances under different combinations of atmosphere model, aerosol model, and solar zenith angle were calculated, and the calculated results are shown as gray lines in Fig. 1(a). The K-means result as the reference solar spectrum is shown as the red line in Fig. 1(a). The correlation coefficients between the reference spectrum and the calculated spectrum under different conditions were analyzed. The probability distribution of the correlation coefficients is shown in Fig. 1(b). The average spectral correlation coefficient is 0.994. The reference spectrum shows high correlation to the calculated spectrum under different conditions, while the irradiance at different solar zenith angles varies over a wide range. Hence, the reference spectrum is considered to represent typical solar spectral features in our solar simulator.

 figure: Fig. 1.

Fig. 1. Spectral distribution (a) and histogram of spectral correlation coefficient (b) between the reference solar spectrum and the calculated solar irradiance under different conditions.

Download Full Size | PDF

2.2 Theory of solar simulation

The proposed solar simulator has an illuminated area of 100 cm x 100 cm and can characterise large-size samples under variable light direction. The spectral range is between 350 nm and 2500 nm which offers the capability of characterising the spectral reflectance of various different materials. According to the requirements of remote sensing simulation, the solar simulator should provide a similar spectral irradiance to sunlight in the laboratory, and control illumination angles to simulate different solar incident angles. The solar simulator is installed on the motion platform and ensures illumination direction always points to the target, to simulate the angular motion track of the sun. The maximum size of the solar simulator is limited to 120 cm × 120 cm because of the limitations of weight and available space. The target plane is at a distance of 3.8 m from the center of the source plane. The solar simulator based on multiple lamps was designed to project a sun-like irradiance on the target surface region. To achieve the abovementioned goals, we designed a solar simulator as shown in Fig. 2.

 figure: Fig. 2.

Fig. 2. Schematic structure of the solar simulator.

Download Full Size | PDF

In the far zone, the extended light source can be treated as a point source with a specific angular intensity distribution. The light intensity distribution can be described by Eq. (1):

$$I(\theta )\textrm{ = }{I_0}{\cos ^m}(\theta )$$
where I (θ) is the light intensity of the light source along the viewing angle θ, and I0 is the light intensity at the normal direction to the source surface. The m value depends on spatial intensity distribution [28], and this is given by Eq. (2)
$$m = \frac{{ - \ln 2}}{{\ln (\cos{\theta _{1/2}})}}$$
where θ1/2 is the viewing angle when the irradiance decreases to the half of the value at the normal direction. If the source is a perfect Lambertian, the m value is 1. If the source is a highly directional emitter, the m value will be large.

And then, the irradiance E of a point on the target plane generated by the source on the source plane, which can be expressed as Eq. (3):

$$E(\theta ,\varphi ,d) = \frac{{{I_0}\textrm{co}{\textrm{s}^m}(\theta )\cos \textrm{(}\varphi \textrm{)}}}{{{d^2}}}$$
where φ is the angle between the target plane normal and the light, and d is the distance between the light source and the target plane.

The spectral irradiance distribution of the light sources on the target plane is derived through the analysis and calculation of the luminous characteristics and the spatial distance of the light sources. The total irradiance on the target plane is calculated by summing each lamp’s light distribution together. When n lamps are used, the expression of total irradiance Esum is given by Eq. (4)

$${E_{\textrm{sum}}} = \sum\limits_{\textrm{i = }1}^\textrm{n} {E({\theta _i},\varphi ,{d_i})} $$

The uniformity of illumination is essential to increase the accuracy of segmenting and matching hyperspectral images. Generally, irradiance values at a large number of observation points over the target plane are calculated, and the uniformity of illumination is evaluated by the abbreviation of the coefficient of variation of root mean square error CV(RMSE), as shown in Eq. (5).

$$CV(RMSE) = \frac{\sigma }{{\overline E }}$$
where $\overline E $ is the average irradiance of all grids on the target plane, and σ is the standard error in irradiance. A smaller CV(RMSE) value represents a higher uniformity.

Spectral similarity has been used to efficiently classify pixels into the known categories in the hyperspectral images [29]. Pearson correlation coefficient provides an appropriate estimate of the similarity between two variables, which is applied to estimate the spectral match performance between the reference solar spectrum and a simulated spectrum in this study. The formula below presents the Pearson coefficient, which has as the numerator the covariance of two variables and as the denominator the product of their standard deviations:

$$r(A,B) = \frac{{\sum\limits_{i = 1}^n {({A_i} - \overline A )({B_i} - \overline B )} }}{{\sqrt {\sum\limits_{i = 1}^n {{{({A_i} - \overline A )}^2}} } \sqrt {\sum\limits_{i = 1}^n {{{({B_i} - \overline B )}^2}} } }}$$
where Ai and Bi are the elements of the intensity vectors representing the spectra under comparison, and $\overline A$ and $\overline B$ are the mean intensity values.

2.3 Proposed two phase genetic algorithm

There are various possibilities for solar simulation solutions depending on the source characteristics, the number of lamps, and the spatial positions. The possible combinations increase exponentially with the number of light sources. The traditional trial-and-error methods are unsuitable for finding a suitable solution due to either inaccurate results or exorbitant computation times. Therefore, we will develop a method to achieve a high spectral match and irradiance uniformity in a large-area solar simulator.

The genetic algorithm, as an evolutionary algorithm, was selected to identify optimal or near-optimal solutions for problems with a large search space [30]. The appeal of genetic algorithms comes from their simplicity and robustness as well as their power to discover good solutions for complex high-dimensional global optimization problems that are very difficult to handle by traditional empirical methods. In addition, the greatest advantage of genetic algorithms that differentiates them from other search techniques is their ability to handle a population of potential solutions, rather than modifying a single point. By performing optimization simultaneously at different regions in the problem domain, a genetic algorithm offers a higher chance of discovering very good results. Genetic algorithms obtain their ability to recognize trends toward optimal solutions by combining the principles of survival of the fittest with randomized information exchange.

To improve the computational efficiency of the search, a two-phase genetic algorithm method is proposed to achieve the desired irradiance level and spectral matching performance. In the first phase, a genetic algorithm is applied for the spectral matching optimization process to find the most feasible solution for the types and number of light sources. The required input data for this phase include the light source intensity, range of number, and spectral characteristics of the individual sources. In the second phase, the genetic algorithm keeps Phase 1 solutions and then maps the light sources to different spatial positions to conduct the spatial optimization process. The required input data for this phase include the spatial position and viewing angle of the light source of the individual sources. The main framework of the proposed two-phase genetic algorithm is presented in Fig. 3.

 figure: Fig. 3.

Fig. 3. A flowchart overview of the two phase genetic algorithm method.

Download Full Size | PDF

The spectral matching optimization solutions and the spatial optimization solution construct a set of complete solutions for a hybrid solar simulator together. In each phase, the algorithm begins with a population of candidate solutions, then continuously modifies them and creates new solutions by genetic operators, i.e., selection, crossover and mutation. After several new generations, the system gradually evolves towards an optimal global solution. However, for solving different optimization problems, the parameters of solution representation, initial population, operators, objective fitness functions, and termination criterion are vary. The process of implementing the two-phase genetic algorithm proposed in this section is presented as follows:

Phase 1: Spectral matching optimization

In the spectral matching optimization phase, the algorithm is performed to select the most appropriate set of light sources to minimize the difference between the theoretical spectrum and the estimated spectrum. In this phase, the type and the number of sources are selected to accurately represent the solar spectrum from a wide variety of lamps available on the market. And the method opens up the possibility to reduce the overall cost of the illumination system and prevent avoidable power consumption.

The algorithm performs the following steps:

  • Step 1.1 (Initialization): Initialize the population by setting the actual problem parameters (e.g., population size M, maximum generation G, crossover probability Pc, and mutation probability Pm) and generate the first-generation population. This study adopts natural number coding the number of different types of source, which makes the construction of the genetic algorithm much simpler.
  • Step 1.2 (Evaluation): Evaluate the fitness function value of each individual in the population. Here, the fitness function should be related to the parameters of the irradiance level and spectral matching.
  • Step 1.3 (Selection): Rate the fitness of each solution and select parents from the initial population. Roulette wheel method is used for selection method.
  • Step 1.4 (Crossover): Create new individuals (offsprings) by combining parts from two randomly selected individuals of the population.
  • Step 1.5 (Mutation): Create a new individual by a small change in a single individual.
  • Step 1.6 (Termination): Repeat steps 1.2 to 1.4 until the maximum number of iterations is reached and determine the solution for the best combination of number and type of different light sources.
Phase 2. Spatial optimization

In the spatial optimization phase, the algorithm keeps the best feasible solution that was found in the first phase, and achieves the accurate selection of the most appropriate position for all sources in order to improve the uncertainties caused by spectral nonuniformity. In this phase, the radiant intensity distributions and position of different LEDs in the lighting scheme are the inputs. The light sources can be irregularly placed anywhere on the lighting scheme. The algorithm repeatedly substitutes different positions solutions of sources to search for the best position in which the source array produces the uniform spectral irradiance.

The algorithm performs the following steps:

  • Step 2.1 (Initialization): Select the best solution of phase 1, construct the initial population for the actual problem, and initialize the population parameters. The permutation encoding scheme is used in the encoding scheme.
  • Step 2.2 (Evaluation): Evaluate the fitness function value of each individual in the population. Here, the fitness function is related to the illumination uniformity parameters.
  • Step 2.3 (Selection): Sort initial population and select parents from the initial population.
  • Step 2.4 (Crossover): Create new individuals (offsprings) by combining parts from two randomly selected individuals of the population.
  • Step 2.5 (Mutation): Create a new individual by a small change in a single individual.
  • Step 2.6 (Termination): Repeat steps 2.2 to 2.4 until the maximum number of iterations is reached and determine the best solution of the configuration.

3. Experiment

3.1 Light source characteristics

According to the characteristics of different types of light sources, we created a commercial sources library for designing a solar simulator, containing metal halide lamps, halogen lamps, high power white LEDs and different kinds of monochromatic LEDs. The spectral distributions of metal-halide lamps, halogen lamps, and white LEDs are shown in the Fig. 4. The metal halide lamp produces well-balanced white light with high brightness, but contributes a multitude of spectral lines in the visible and infrared spectral regions. For example, there are some significant peaks of the range of 540 -600 nm, which tend to mismatch the solar spectrum. Halogen lamps serve as broad-spectrum light sources providing sufficient irradiance in the visible and infrared ranges. White LEDs contribute most of the required optical power for the 420 nm to 480 nm and 500 nm to 700 nm ranges in the visible spectrum range. In addition, we researched 31 kinds of monochromatic LEDs for covering the spectral gap in the 350-1000 nm solar spectral region. The relative spectral power distributions of monochromatic LEDs are shown in Fig. 5.

 figure: Fig. 4.

Fig. 4. Spectral distribution of a metal halide lamp, halogen lamp and white LED.

Download Full Size | PDF

 figure: Fig. 5.

Fig. 5. The relative spectral power distributions of 31 kinds of monochromatic LEDs.

Download Full Size | PDF

Metal halide lamps are based on advanced spotlighting technology, providing a 10-degree angle beam of light, as shown in Fig. 6(a). Halogen lamps with an integrated reflector could provide light at a similar beam angle, as shown in Fig. 6(b). However, LEDs typically emit light at a 120-degree angle, which makes it difficult to meet the consistency of the spatial intensity distribution. Here, the total internal reflector (TIR) lens is designed to ensure high beam directionality and increase collection efficiency, as shown in Fig. 6(c). The TIR lens can be placed directly on top of each LED chip to send out a concentrated beam within 10 degrees. The spatial luminous intensity distributions of a metal halide lamp, a halogen lamp, a LED with a TIR lens are shown in Fig. 6(d).The luminous intensity patterns of different light sources are basically consistent, which ensures the consistency of the emitted light angle of the solar simulator.

 figure: Fig. 6.

Fig. 6. (a) Photograph of the metal halide lamp; (b) photograph of the halogen lamps with an integrated reflector; (c) photograph of the TIR lens; (d) the spatial intensity distributions of the three types of sources.

Download Full Size | PDF

In the next section, we will choose the most appropriate scheme from the light source library to perform the simulation of solar spectral irradiance.

3.2 Spectrum match optimization procedure

The first criterion in our design approach was to develop to match the emulated light spectrum to the objective solar spectrum. By comparing the characteristics of different types of light sources, a metal halide lamp was selected as the basic light source. Halogen lamps and different LEDs were introduced to meet the requirement of the spectral match. However, the increasing number of lamps will lead to higher costs and more complexity of the design. Thus, a more accurate methodology is crucial to reach a satisfactory solution from the seemingly endless number of potential combinations.

In this study, an automated methodology for choosing the appropriate sources was designed to find the optimum solution for achieving the desired irradiance and spectral performance with fewer light sources. The number and spectral distribution of light sources are provided as input parameters. We use the natural number coding method method to encode the combination of different light sources. Individual solutions of the initial population are generated by the random method, which can prevent the optimal solution from falling into a local optimum in the optimization process. A roulette operator is selected to accelerate the convergence of individual solutions and ensure the stability of genetic inheritance among the population. Two objective functions were used in this study to achieve the desired irradiance and maximize the spectral match. To reduce the overall cost of the illumination system, the total number of light sources in the configuration was used for the constrained problem. The objective function is described as follows:

$$\begin{array}{c} {\textrm{Minimize}}\,{f_1} = {\omega _1} \cdot (\overline E - {E_{obj}}) + {\omega _2} \cdot \frac{1}{{(r - {r_{obj}})}}\\ {\textrm{Subject}}\textrm{ }to\,\overline E \ge {E_{obj}}\,{\textrm{and}}\,r \ge {r_{obj}} \end{array}$$
where, ${\omega _1}$ and ${\omega _2}$ are weight coefficients, ${E_{obj}}$ is the desired irradiance, and $\overline E$ is the average irradiance of the target area in the design solution, r is Pearson’s correlation coefficient between the reference spectrum and calculated spectrum, and ${r_{obj}}$ is the low boundary of the Pearson’s correlation coefficient.

The crossover probability was set to 0.75, and mutation probability, was set to 0.1. A predefined iteration number of 30000 was used as the termination rule in this study. The optimization routine determined the best solution for selecting the number and type of light sources. Table 1 summarizes the characteristics of the sources selected for the solar simulator design.

Tables Icon

Table 1. Characteristics of the sources selected for the solar simulator design.

View Table | View all tables in this article

The solution selected 25 types of light sources from the previous light source database. The simulation spectrum was calculated using the spectral irradiance of each type of source multiplied by its number. The simulated theoretical spectrum of the simulator is shown in Fig. 7, together with the reference solar spectrum. The metal halide lamp served contributed 20% of the power in the full spectral range, while the halogen lamps also provided 24.6% of the power in the visible and near-infrared wavelength range. Twenty-nine white LEDs served as broad-spectrum light sources to contribute 16.4% of the power in the visible range, which can be employed to save a certain amount of monochromatic LEDs in the visible range of the spectrum. In the optimal combination, 4 species of monochromatic LEDs are removed for the source data set; the removed monochromatic LEDs are 404 nm, 574 nm, 586 nm, and 590 nm. Twenty-two kinds of monochromatic LEDs were selected to complement the remaining spectral region. Finally, the correlation coefficient between the simulated spectrum and the reference solar spectrum is 0.916, which shows a good spectral match.

 figure: Fig. 7.

Fig. 7. The normalized spectral profiles of the reference solar spectrum and simulated spectrum.

Download Full Size | PDF

3.3 Light distribution optimization procedure

According to the modular design idea, rectangular module geometry can be tessellated to provide a large-area illumination. An efficient, compact, solar simulator system can be constructed from multiple rectangular units. Since light sources are organized in the same configuration, it is possible to achieve irradiance variation by selectively turning on the illumination modules without changing the spectral characteristics. In this study, 25 illumination modular units were used to achieve the total irradiance requirement over a large-scale target surface. However, the arrangements of a variety of lamps directly affect the uniformity of spectral irradiance on the target plane. In particular, the spectral irradiance uniformity issue can arise when the area of illumination increases.

The second procedure in our design approach is to find an optimal layout for reducing the spectral irradiance nonuniformity. For each independent modular unit, one separate metal halide lamp was assembled with 4 halogen lamps and 128 LEDs. The metal halide lamp is mounted in the center of the module, and the halogen lamps are arranged in a symmetrical (diagonal) pattern surrounding the metal halide lamp. For each illumination modular unit, 128 high-power LEDs are organized into four groups. The outer size (W × H) of each cluster of LEDs array is 60 mm × 120 mm. The LED groups are positioned at the sides of the metal halide lamp, and the distance from the center of metal halide lamp is 90 mm. We analyzed metal halide lamps arrays and halogen lamps arrays, and then focused on the LED array configuration. The arrangements of metal halide lamps and halogen lamps on 5 × 5 array rectangular configurations are as shown in Fig. 8. The nonuniformity of the metal halide lamp array is calculated as 13.71%, while the nonuniformity of the halogen lamp array is calculated as 14.86%.

 figure: Fig. 8.

Fig. 8. The arrangement of metal halide lamp array (a) and halogen lamp array (b).

Download Full Size | PDF

However, different positions of LEDs will cause different luminance distributions on the receiving plane, especially for asymmetric distribution structures. The LED arrays may produce large shifts in the irradiation pattern on the target surface. Finding the optimum positions of over 600 LEDs of 23 different wavelengths is not a trivial task due to the lots of possible combinations. Hence, an efficient and accurate method was developed to minimize the difference of the spectral irradiance map between metal halide lamp array and the most LEDs array. A matrix containing the number of LEDs needed per wavelength is the main input variable. The algorithm initially assigns randomly each LED to a position. The gene consists of a sequence of numbers equal to the total number of LEDs. Each number can take any value between 1 and the maximum number of wavelengths, 23 in this case. Therefore, each gene will be a sequence of 128 numbers, and each corresponding to a position can be used to place one of 23 types of LED. Each number i will appear in the gene sequence Xi times, where Xi is the number of LEDs for the wavelength i. Then a number of permutations of these positions is calculated and forms the initial population. The random method is adopted for initializing the population so that better candidate solutions can be identified in the search for the global optimal solution. The evaluation function of the illumination system was established by taking the illumination uniformity of metal halide lamps as the main evaluation index and comprehensively considering the illumination pattern. To achieve a highly uniform of spectral irradiance, the least-squares is used to minimize the objective function of the illumination system, as follows:

$$Minimum\,{f_2} = {\sum\limits_{i = 1}^{i = n} {({C{V_{MHP}} - C{V_{LED}}({X_i})} )} ^2}$$
where, Xi is the i type series of LEDs, and i is from 1 to 23;

The irradiance distribution of LED arrays with optimization configuration and the sequential configuration were analyzed. In the optimal configuration, the most appropriate positions for LED sources have been achieved by the created algorithm, as shown in Fig. 9(a). The nonuniformity results of different types of LEDs for the solar simulator were calculated, as shown in Fig. 9(b). Ninety-one percent of the total number of LEDs had a nonuniformity of illumination of less than 18%. A few LEDs have a higher nonuniformity due to non-central arrangement of these LEDs. In the sequential configuration, LEDs are placed in different position nodes in sequence, as shown in Fig. 10(a). Then spectral irradiance distributions of different types of LEDs were computed, and the non-uniformity parameters of different types of LEDs were given in the Fig. 10(b). There are 78.9% of the total number of LEDs with non-uniformity of illumination less than 18%. Obviously, that the improved algorithm is efficient and has a great contribution to the spectral uniformity of the LED array. The optimal configuration is selected to improve the uniformity of the spectral irradiance distribution for the hybrid solar simulator.

 figure: Fig. 9.

Fig. 9. (a) Source position in the optimal configuration and (b) the nonuniformity of each type of LEDs.

Download Full Size | PDF

 figure: Fig. 10.

Fig. 10. (a) Source position in the sequence configuration and (b) the nonuniformity of each type of LEDs.

Download Full Size | PDF

The optical performance of the hybrid solar simulator can be simulated and verified using ZEMAX simulation software. The spectral and spatial emission profiles for each lamp were used as the input for the ray tracing model. The lamps are represented as extended light sources. Samples of 10^6 rays per lamp are used in the simulation. The rays are generated randomly emanating from the emitting surface of a lamp. The Monte Carlo ray-tracing approach is used to simulate the lamp’s emitted light rays that reach the target plane. The resolution of the target area is 100 × 100 and the irradiance is recorded every cm2. The overall spectral intensity distribution of the solar simulator is obtained by summering the spectral irradiance distributions of different types of light sources. The irradiance distributions for the hybrid solar simulator were simulated at typical zenith angles. Figures 11(a)–11(d) show the irradiance maps over the test plane at the zenith angle of 0°, 10°, 40°, and 70°. The illumination uniformity are 14.99%, 14.82%, 13.1%, and 12.92% respectively, and average irradiance are 724.5 W/m2, 723.8 W/m2, 589 W/m2, and 276.6 W/m2, respectively. The simulation results show that solar simulator achieves the uniform irradiance distribution over the test plane.

 figure: Fig. 11.

Fig. 11. The simulated irradiance distributions of the solar simulator at different zenith angles: (a) 0° zenith angle, (b) 10° zenith angle, (c) 40° zenith angle, (d) 70° zenith angle.

Download Full Size | PDF

4. Results and discussion

4.1 System manufacturing

We conducted experiments to check the validity of the simulation results. In an illumination modular unit, a metal halide lamp and four halogen lamps are installed with custom bases. The 128 LEDs are divided into four groups. The 32 LEDs of each group are mounted on a metal core printed circuit board (MCPCB), which dimension of 12 cm × 6 cm. However, high power LEDs usually generate a lot of heat during operation, which influences the intensity and spectral properties. Therefore, thermal management is necessary to achieve good temporal stability. High-power LED light sources generally use active cooling or passive cooling for heat dissipation. To verify the performances of the two cooling methods, a set of comparative experiments were conducted. An infrared imager was used to measure the temperature of the LED array. In the case of passive cooling, the MCPCB was mounted on a finned heat exchange system made of an extruded metal heat sink (12 cm × 6 cm × 4 cm). In Fig. 12(a), the temperature of the LED array was increasing and continued to increase to 92.3 °C. In the case of active cooling, a cooling fan (5 cm × 5 cm, 24 V 0.15 A) was combined with the heat sink to improve the efficiency of air heat exchange. The details in Fig. 12(b) show that the LED chip temperature remained below 47 °C during operation. Consequently, the solution of active cooling should be used to offer higher heat dissipation for the illumination unit.

 figure: Fig. 12.

Fig. 12. (a) Infrared image of the PCB with LEDs under passive cooling; (b) Infrared image of the PCB with LEDs under active cooling.

Download Full Size | PDF

The solar simulator was designed with attention to ease of setup, calibration, and testing. Each illumination modular unit including power supplies was independently assembled on an aluminum rectangular plate. The support structure of the light source was provided by an aluminum alloy plate with sufficient rigidity and light weight. The 25 illumination units of the solar simulator were assembled on an aluminum rectangular structure, and then fixed on the frame of motion platform. The overall size of the solar simulator is 1.2 m × 1.2 m × 0.24 m. The prototype of the solar simulator is shown in Fig. 13. Each illumination unit can work independently by the nominal 220-V AC application. When all illumination units are turned on, the maximum power of the solar simulator can reach 13 kW. This solar simulator can provide a suitable illumination that similarity to direct solar spectral irradiance for remote sensing physical simulation in the laboratory.

 figure: Fig. 13.

Fig. 13. Photograph of the fabricated solar simulator.

Download Full Size | PDF

4.2 Experiment and analysis

To determine the performance of the solar simulator, we measured three parameters: spectral match, irradiance spatial nonuniformity, and temporal instability.

4.2.1 Spectral match

The spectral distribution of the solar simulator was measured using a calibrated spectrometer (Analytical Spectral Devices, CO, USA) with a measurement range of 350-2500 nm. The measurements were conducted in a dark laboratory environment. The solar simulator illuminated the target surface with a 10° zenith angle from a distance of 3.8 m. The spectra were collected from the central area of the target surface. The final spectrum measurement was attained by averaging the five representative spectra. The measured spectrum of the hybrid solar simulator was compared to reference sunlight and the spectrum of metal halide lamps, as shown in Fig. 14. Compared to the solar simulator based on metal halide lamps, the correlation coefficient between the spectrum of metal halide lamps and the reference spectrum just only was 0.65. The results show that the correlation coefficient between the measured spectrum and the reference spectrum was calculated as 0.90.The percentages of VIS, NIR, and SWIR regions relative to the total integrated irradiance in the measured spectrum are 57.4%, 32.3%, and 10.3%, respectively, while the percentages in the reference spectrum are 50.6%, 31.7%, and 17.7%, respectively. These measurement results were nearly consistent with the simulation results, verifying the effectiveness of this design method. The hybrid solar simulator provides a better spectral match for the natural sunlight in the wavelength range from 350 nm to 2500 nm.

 figure: Fig. 14.

Fig. 14. Comparison the measured spectrum to the reference spectrum and metal halide lamps.

Download Full Size | PDF

4.2.2 Irradiance nonuniformity

The nonuniformity (NU) is a performance parameter that is used to determine the variation of irradiance over the target surface. A professional pyranometer (Solar Light's Model PMA2145) was used to measure the irradiance distribution on the test surface, which has a very broad spectral sensitivity, exceptional flatness, and excellent long term stability. The total measured area is 100 × 100 cm2. The target area was equally divided into 100 squares with a grid spacing of 10 cm. The irradiance was measured at each square on the target plane. Due to the influence of the mechanical structure, the solar simulator could move between the 10° zenith angle and the 70° zenith angle. The solar simulator’s uniformity was evaluated at the minimum zenith angle (10°), the minimum zenith angle (40°) and the maximum zenith angle (70°). The experimental results of irradiance distributions of the solar simulator at three zenith angles are shown in Fig. 15.

 figure: Fig. 15.

Fig. 15. Light intensity of the solar simulator at different zenith angle: (a) 10° zenith angle; (b) 40° zenith angle; (c) 70° degree zenith angle.

Download Full Size | PDF

From the resulting irradiance map of the test plane, the parameter spatial nonuniformity can be calculated. At solar zenith angle of 10°, the arithmetic mean of the measured light intensity values is 726 W/m2, which is close to the intensity of direct sunlight on the Earth's surface. The irradiance distribution is relatively homogenous within the inspection range, and irradiance non-uniformity can be calculated to 15.74%. When the solar simulator is moved to the position of the solar zenith angle of at°, the average irradiance on the test plane is 565 W/m2, and the value of the nonuniformity was equal to 15.86%. When the test plane is illuminated by the solar simulator at the solar zenith angle at 70°, the average irradiance is 274 W/m2 and the non-uniformity was equal to 16.74%. Table 2 shows the results of average irradiance and nonuniformity of the optical simulation and actual measurement at different zenith angles. It can be seen that the irradiance of the test plane decreases as solar zenith angle increases, and the measured results are close to the simulated results. The errors between the actual measurements and the simulated measurement may come from measurement errors and systematic errors. Moreover, the hybrid solar simulator also leaves a considerable potential that the illumination irradiance at different solar zenith angles could be accurately obtained by adjusting the number of illumination units.

Tables Icon

Table 2. The results of average irradiance and nonuniformity of the simulation and measurement for the solar simulator at different zenith angle

View Table | View all tables in this article

4.2.3 Instability of irradiance

Temporal instability of irradiance is calculated by (5) with reference to the time period of interest. The solar simulator was warm up and reach steady operation. Considering the test requirements, we have measured irradiance values in 30 minutes using the calibrated spectrometer. The results of the total radiance are as shown in Fig. 16. The result shows that temporal instability of the solar simulator radiance was 3.38%.

 figure: Fig. 16.

Fig. 16. The irradiance instability results of solar simulator.

Download Full Size | PDF

5. Conclusion

In this paper, we discuss the design and fabrication of a large-area hybrid solar simulator for remote sensing imaging simulation. A design procedure based on a two-phase genetic algorithm has been proposed to automatically select the best configuration of multiple lamps and optimize the arrangement of the source array. This solar simulator was designed to achieve a high spectral match and spatial uniformity and simulate the variations of different solar zenith angles. The experimental results showed the correlation coefficient between the measured spectrum and the reference spectrum could reach 0.90 in the wide wavelength range 350 nm to 2500 nm. The average irradiance over an area of 100 × 100 cm2 could achieve 762 W/m2 at 10° solar zenith angles, while the non-uniformity values was 15.74%.When the solar simulator moved to 70° solar zenith angles, the average irradiances was 274 W/m2 and the nonuniformity value was 16.74%.The temporal instability of irradiance was lower than 3.38%. Our method achieves good accuracy in terms of both spectral match and spatial uniformity. Additionally, the modular structure of this design has the great advantage that the irradiance can be accurately controlled by adjusting the number of illumination modules according to the test requirements. This design opens up the possibility of achieving the desired spectral irradiance, high-uniformity, and large-area solar simulation, without filters and complex optical systems. This hybrid solar simulator can provide a valuable novel experimental platform for remote sensing simulation in the laboratory. In future, we will apply the solar simulator to more interest application, such as hyperspectral imaging, vegetation index evaluation, and the reflectance feature measurement, etc.

Acknowledgments

The authors would like to thank the support of Postdoctoral research project of Qingdao Research Institute of Beihang University and Qingdao Postdoctoral Application Research Project.

Disclosures

The authors declare no conflicts of interest.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

References

1. H. Zhao, B. Cui, G. Jia, G. Jia, N. Li, X. Li, X. Li, F. Gan, and J. Yu, “Acquirement of anisotropy reflectance with the multi-directional hyperspectral remote sensing simulation facility (MHSRS2F),” Opt. Express 27(20), 28760–28781 (2019). [CrossRef]  

2. S. Mertens, L. Verbraeken, H. Sprenger, K. Demuynck, K. Maleux, B. Cannoot, J. De Block, S. Maere, H. Nelissen, G. Bonaventure, S. J. Crafts-Brandner, J. T. Vogel, W. Bruce, D. Inzé, and N. Wuyts, “Proximal hyperspectral imaging detects diurnal and drought-induced changes in maize physiology,” Front. Plant Sci. 12, 640914 (2021). [CrossRef]  

3. M. S. Roxy, V. B. Sumithranand, and G. Renuka, “Variability of soil moisture and its relationship with surface albedo and soil thermal diffusivity at Astronomical Observatory, Thiruvananthapuram, south Kerala,” J. Earth Syst. Sci. 119(4), 507–517 (2010). [CrossRef]  

4. Y. Li, Z. Jiao, K. Zhao, Y. Dong, Y. Zhou, Y. Zeng, H. Xu, X. Zhang, T. Hu, and L. Cui, “Influence of varying solar zenith angles on land surface phenology derived from vegetation indices: a case study in the harvard forest,” Remote Sens. 13(20), 4126 (2021). [CrossRef]  

5. X. Li, J. Chen, W. Lipiński, Y. Dai, and C.-H. Wang, “A 28 kWe multi-source high-flux solar simulator: Design, characterization, and modeling,” Sol. Energy 211, 569–583 (2020). [CrossRef]  

6. H. Zhao, B. Cui, G. Jia, X. Li, C. Zhang, and X. Zhang, “A “Skylight” simulator for HWIL simulation of hyperspectral remote sensing,” Sensors 17(12), 2829 (2017). [CrossRef]  

7. R. Cortés-Severino, C. Cárdenas-Bravo, R. Barraza, A. Sánchez-Squella, P. V. Lefort, and F. Castillo-Burns, “Optimal design and experimental test of a solar simulator for solar photovoltaic modules,” Energy Sci. Eng. 9, 2514–2528 (2021). [CrossRef]  

8. G. Levêque, R. Bader, W. Lipiński, and S. Haussener, “High-flux optical systems for solar thermochemistry,” Sol. Energy 156, 133–148 (2017). [CrossRef]  

9. J. Pottas, L. Li, M. Habib, C.-H. Wang, J. Coventry, and W. Lipiński, “Optical alignment and radiative flux characterization of a multi-source high-flux solar simulator,” Sol. Energy 236, 434–444 (2022). [CrossRef]  

10. M. Tawfik, X. Tonnellier, and C. Sansom, “Light source selection for a solar simulator for thermal applications: A review,” Renewable Sustainable Energy Rev. 90, 802–813 (2018). [CrossRef]  

11. R. A. Salam, M. M. Munir, T. Warsahemas, C. Saputra, H. Latief, and K. Khairurrijal, “A simple solar simulator with highly stable controlled irradiance for solar panel characterization,” Meas. Control 52(3-4), 159–168 (2019). [CrossRef]  

12. R. Bader, S. Haussener, and W. Lipiński, “Optical design of multisource high-flux solar simulators,” J. Sol. Energy Eng. 137(2), 021012 (2015). [CrossRef]  

13. W. Wang, L. Aichmayer, B. Laumert, and T. Fransson, “Design and validation of a low-cost high-flux solar simulator using fresnel lens concentrators,” Energy Procedia 49, 2221–2230 (2014). [CrossRef]  

14. R. Bader, G. Levêque, S. Haussener, and W. Lipiński, “High-flux solar simulator technology,” Optica Publishing Group, SoM3C.3 (2016).

15. G. Levêque, R. Bader, W. Lipiński, and S. Haussener, “Experimental and numerical characterization of a new 45 kWel multisource high-flux solar simulator,” Opt. Express 24(22), A1360–A1373 (2016). [CrossRef]  

16. M. Hamady, G. G. Lister, and G. Zissis, “Calculations of visible radiation in electrodeless HID lamps,” Light. Res. Technol. 48, 502–515 (2016). [CrossRef]  

17. E. Yandri, “Uniformity characteristic and calibration of simple low cost compact halogen solar simulator for indoor experiments,” Int. J. Low-Carbon Technol. 13(3), 218–230 (2018). [CrossRef]  

18. S. Kohraku and K. Kurokawa, “A fundamental experiment for discrete-wavelength LED solar simulator,” Sol. Energy Mater. Sol. Cells 90(18-19), 3364–3370 (2006). [CrossRef]  

19. Y. Tsuno, K. Kamisako, and K. Kurokawa, “New generation of PV module rating by LED solar simulator - A novel approach and its capabilities,” in 2008 33rd IEEE Photovoltaic Specialists Conference (IEEE, 2008), pp. 1–5.

20. J. R. Lorentzen, D. A. Scheiman, R. Hoheisel, R. J. Walters, and P. P. Jenkins, “Impact of solar simulator spectrum on multi-junction solar cell measurement,” in 2015 IEEE 42nd Photovoltaic Specialist Conference (PVSC) (IEEE, 2015), pp. 1–3.

21. G. Grandi, A. Ienina, and M. Bardhi, “Effective low-cost hybrid LED-halogen solar simulator,” IEEE Trans. Ind. Appl. 50(5), 3055–3064 (2014). [CrossRef]  

22. A. Y. Al-Ahmad, J. Holdsworth, B. Vaughan, G. Sharafutdinova, X. Zhou, W. J. Belcher, and P. C. Dastoor, “Modular LED arrays for large area solar simulation,” Prog. Photovoltaics Res. Appl. 27(2), 179–189 (2019). [CrossRef]  

23. A. Novičkovas, A. Baguckis, A. Mekys, and V. Tamošiūnas, “Compact light-emitting diode-based AAA class solar simulator: design and application peculiarities,” IEEE J. Photovoltaics 5(4), 1137–1142 (2015). [CrossRef]  

24. F. Zhang and G. Zhou, “Estimation of vegetation water content using hyperspectral vegetation indices: a comparison of crop water indicators in response to water stress treatments for summer maize,” BMC Ecol. 19(1), 18 (2019). [CrossRef]  

25. A. S. Olpenda, K. Stereńczak, and K. Będkowski, “Modeling solar radiation in the forest using remote sensing data: a review of approaches and opportunities,” Remote Sens. 10(5), 694 (2018). [CrossRef]  

26. D. Nowak, L. Vuilleumier, C. N. Long, and A. Ohmura, “Solar irradiance computations compared with observations at the Baseline Surface Radiation Network Payerne site,” J. Geophys. Res. Atmos. 113(D14), D14206 (2008). [CrossRef]  

27. G. Tzortzis and A. Likas, “The MinMax k-Means clustering algorithm,” Pattern Recognit. 47(7), 2505–2516 (2014). [CrossRef]  

28. Z. Su, D. Xue, and Z. Ji, “Designing LED array for uniform illumination distribution by simulated annealing algorithm,” Opt. Express 20(S6), A843–A855 (2012). [CrossRef]  

29. J. N. Sweet, “The spectral similarity scale and its application to the classification of hyperspectral remote sensing data,” in IEEE Workshop on Advances in Techniques for Analysis of Remotely Sensed Data, 2003 (IEEE, 2003), pp. 92–99.

30. F. T. Silva, M. X. Silva, and J. C. Belchior, “A new genetic algorithm approach applied to atomic and molecular cluster studies,” Front. Chem. 7, 707 (2019). [CrossRef]  

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

Cited By

Optica participates in Crossref's Cited-By Linking service. Citing articles from Optica Publishing Group journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (16)
Fig. 1.
Fig. 1. Spectral distribution (a) and histogram of spectral correlation coefficient (b) between the reference solar spectrum and the calculated solar irradiance under different conditions.

View in Article | Download Full Size | PDF

Fig. 2.
Fig. 2. Schematic structure of the solar simulator.

View in Article | Download Full Size | PDF

Fig. 3.
Fig. 3. A flowchart overview of the two phase genetic algorithm method.

View in Article | Download Full Size | PDF

Fig. 4.
Fig. 4. Spectral distribution of a metal halide lamp, halogen lamp and white LED.

View in Article | Download Full Size | PDF

Fig. 5.
Fig. 5. The relative spectral power distributions of 31 kinds of monochromatic LEDs.

View in Article | Download Full Size | PDF

Fig. 6.
Fig. 6. (a) Photograph of the metal halide lamp; (b) photograph of the halogen lamps with an integrated reflector; (c) photograph of the TIR lens; (d) the spatial intensity distributions of the three types of sources.

View in Article | Download Full Size | PDF

Fig. 7.
Fig. 7. The normalized spectral profiles of the reference solar spectrum and simulated spectrum.

View in Article | Download Full Size | PDF

Fig. 8.
Fig. 8. The arrangement of metal halide lamp array (a) and halogen lamp array (b).

View in Article | Download Full Size | PDF

Fig. 9.
Fig. 9. (a) Source position in the optimal configuration and (b) the nonuniformity of each type of LEDs.

View in Article | Download Full Size | PDF

Fig. 10.
Fig. 10. (a) Source position in the sequence configuration and (b) the nonuniformity of each type of LEDs.

View in Article | Download Full Size | PDF

Fig. 11.
Fig. 11. The simulated irradiance distributions of the solar simulator at different zenith angles: (a) 0° zenith angle, (b) 10° zenith angle, (c) 40° zenith angle, (d) 70° zenith angle.

View in Article | Download Full Size | PDF

Fig. 12.
Fig. 12. (a) Infrared image of the PCB with LEDs under passive cooling; (b) Infrared image of the PCB with LEDs under active cooling.

View in Article | Download Full Size | PDF

Fig. 13.
Fig. 13. Photograph of the fabricated solar simulator.

View in Article | Download Full Size | PDF

Fig. 14.
Fig. 14. Comparison the measured spectrum to the reference spectrum and metal halide lamps.

View in Article | Download Full Size | PDF

Fig. 15.
Fig. 15. Light intensity of the solar simulator at different zenith angle: (a) 10° zenith angle; (b) 40° zenith angle; (c) 70° degree zenith angle.

View in Article | Download Full Size | PDF

Fig. 16.
Fig. 16. The irradiance instability results of solar simulator.

View in Article | Download Full Size | PDF

Tables (2)

Tables Icon

Table 1. Characteristics of the sources selected for the solar simulator design.

View Table | View all tables in this article
Tables Icon

Table 2. The results of average irradiance and nonuniformity of the simulation and measurement for the solar simulator at different zenith angle

View Table | View all tables in this article

Equations (8)

Equations on this page are rendered with MathJax. Learn more.

I ( θ )  =  I 0 cos m ( θ )
m = − ln ⁡ 2 ln ⁡ ( cos ⁡ θ 1 / 2 )
E ( θ , φ , d ) = I 0 co s m ( θ ) cos ⁡ ( φ ) d 2
E sum = ∑ i =  1 n E ( θ i , φ , d i )
C V ( R M S E ) = σ E ¯
r ( A , B ) = ∑ i = 1 n ( A i − A ¯ ) ( B i − B ¯ ) ∑ i = 1 n ( A i − A ¯ ) 2 ∑ i = 1 n ( B i − B ¯ ) 2
Minimize f 1 = ω 1 ⋅ ( E ¯ − E o b j ) + ω 2 ⋅ 1 ( r − r o b j ) Subject   t o E ¯ ≥ E o b j and r ≥ r o b j
M i n i m u m f 2 = ∑ i = 1 i = n ( C V M H P − C V L E D ( X i ) ) 2
Select as filters
Select Topics Cancel
Top
       
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.
Privacy | Terms of Use