Comparative performance of different solar still configurations

Abstract

Keywords: Black cloth; DNN; Fins; Solar still; thermal performance; water collected

Introduction

Researchers around the world are increasing their focus on developing sustainable and renewable energy, and thereby transforming the world’s energy landscape. Of the various renewable energy solutions, solar energy has gained significant attention due to its capability to address the growing energy demand along with mitigating the environmental impact [1]. Though solar energy is employed for numerous applications, solar desalination plays a vital

role, as it provides a viable solution for addressing the acute water shortages in arid regions having limited or no freshwater options [2]. Solar stills, a subset of solar desalination, demonstrated significant success in harnessing solar energy for water purification [3]. Basin type solar stills evaporate the saline water by harnessing the solar energy, followed by the condensation of vapor to yield freshwater [4]. Earlier researchers attempted different modifications, materials, and varied process

*Corresponding author. *E-mail address: ssvcdm@gmail.com This paper was recommended for publication in revised form by Editor-in-Chief Ahmet Selim Dalkilic Published by Yıldız Technical University Press, İstanbul, Turkey Yıldız Technical University. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

parameters to augment the performance of the solar stills. Ensafisoroor et al. introduced steps and placed sponges on the basin solar still and successfully increased the water production [5]. Likewise, Hansen et al. used flat, grooved, and fin-shaped absorber plates in an inclined solar still to recover waste heat, resulting in higher water productivity [6]. Goshayeshi and Safaei studied the effect of varying the glass cover’s inclination angle (25°, 27.5°, 30°, 32.5°, and 35°) in stepped solar stills and recommended an optimal angle of 32.5° for improved performance [7]. In a related effort, Omara et al. developed a pyramid-shaped solar still with convex cylinders and nanocomposite-based dish absorber plates to achieve greater efficiency [8]. Ahamed et al. attempted phase change material encapsulated fins and thereby successfully enhanced the thermal performance of the basin solar still [9]. Meanwhile, Kabeel et al. employed v-corrugated absorber plates in a tubular solar still and reported improved performance. In addition, they also developed a mathematical model to replicate the operation of the solar still as well [10]. On the other hand, Hameed et al. performed an experimental study on a single slope solar still and correlated the results with the numerical simulation performed using COMSOL. According to their results, both numerical and experimental outcomes are in good agreement with each other [11]. In 2021, Jobrane et al. summarized the research carried out by other researchers on wick-type solar stills, involving variations in geometrical designs, wick arrangements, and integrated heating systems [12]. Few researchers introduced machine learning on predicting the performance of solar stills. Wang et al. successfully applied the Bayesian optimization algorithm to predict the performance of tubular solar stills [13]. Moustafa et al. utilized an artificial neural network (ANN) to forecast the thermal efficiency and water production of solar stills [14]. Similarly, Bahiraei et al. used a hybrid ANFIS model and a PSO-enhanced neural network to predict energy efficiency with improved accuracy [15].

It is inferred that the published literature focused on varying the absorbing surface, condensation enhancement techniques, and geometric modifications to augment the overall performance of a solar still. Likewise, different types of absorber plates and transparent covers are attempted to improve the heat absorption and to curtail the heat losses [16]. In addition, the effect of climatic conditions such as solar radiation [17, 18], ambient temperature [19], and wind speed [20] on the performance of basin solar stills is reported. The present study proposes the integration of fins and black cotton cloth in a pyramid solar still to attain an improved heat transfer and absorption, leading to higher water yield. The use of the machine learning technique (DNN) for predicting the performance of the solar still provides scope for future improvements in system design and optimization with minimum experimentation.

Materials And Methods

In this study, three different types of solar stills viz., a conventional solar stills (CSS), a pyramid solar still with fins (PSSF), and a pyramid solar still with fins covered with black cloth (PSSFB), were fabricated using 18-gauge alloy sheets, having a uniform basin area of 1 m². These were designed and uniformly tested at the Department of Mechanical Engineering, Annamalai University, Tamilnadu, India. The dimensions of the stills were 110 cm in length and 110 cm in breadth. The CSS serves as the reference, whose basin is coated black, covered with a transparent 4 mm thick glass, having transitivity of 88%, and inclined at 11o (based on the location) to maximize the solar exposure. Water is filled to a depth of 1 cm for effective operation. The edges of the transparent glass cover are fixed with a rubber gasket to curtail the losses. The pyramid solar still (Fig. 1: PSSF) retains the basic components of CSS apart from having the pyramid shaped glass cover, fabricated using four glasses having an equal slope. The covers were attached and sealed

Figure 2. Pyramid shaped solar still with fins covered by black cloth.

using silicone sealant. The metallic fins in the pyramid shaped absorber plate are also coated with black paint and positioned at 90o. In the third setup (PSSFB), a layer of black cloth was employed to cover the fins (Fig. 2). The condensate was collected in a trough located at the base of the glass cover. Thermocouples were positioned at various locations of the solar stills to measure the temperatures of the absorber and glass plates and water. Temperature data were logged using a twelve-channel temperature indicator. A digital thermometer was used to measure the ambient temperature, while the Solar HT kit with a PV204 Solarimeter with an accuracy of ±1 W/ m² was utilized to record the solar intensity. The distilled water was collected in a transparent beaker. Experiments were conducted between 09:00 a.m. and 4:00 p.m. over 25 different days from March to May 2024. The variation in glass, absorber, water, and ambient temperatures; solar radiation; and water yield were recorded at an equal interval of 30 minutes. Subsequent to the experimentation, the obtained data were utilized to develop an artificial intelligence model

employing a deep neural network, which is described below. Of the available data, 70% was used for training the model, and the remaining 30% was equally used for validation and testing on a Windows computer. Deep Neural Network A deep neural network is a complex model based on machine learning techniques that comprises several layers interconnected by neurons (Fig. 3). It is employed to learn complicated patterns within large datasets automatically [21]. There are several neurons within each layer of a DNN that transform the input data and refine features in order to make an accurate prediction. Moreover, the model is equipped with advanced optimization techniques and a host of activation functions, hence very effective in predictive analytics [22]. The depth and architecture are critical in DNN, as this generalizes from training data onto unseen examples and thus drives actual performance in deployment. The accuracy of predicting the responses (temperatures and water collection), with respect to training and testing data, is determined with the aid of three performance

metrics: a) Mean Absolute Error (MAE), b) the coefficient of determination (R²), and c) Mean Absolute Percentage Error (MAPE). MAE represents the average magnitude of errors between predicted and actual values. R² indicates the proportion of variance in the actual data explained by the model. MAPE expresses the error as a percentage by averaging the absolute percentage differences between predicted and actual values. These metrics quantify the deviation between predicted and experimental results. The attempted performance measures are calculated by: [23]. (1)

Results And Discussion

Variation in Solar Intensity Over Time The solar intensity (power received from the sun per unit area) is a critical factor influencing the evaporation and condensation process and thereby dictating the overall efficiency and performance of solar stills. Figure 4 illustrates the fluctuations in solar radiation over the daytime. Further, it is observed that a uniform pattern in solar intensity is observed. The solar intensity varies throughout the day due to the rotation of the earth and the position of the sun in the sky [24]. In the early morning hours, the sunlight reaches the earth at a shallow angle, and hence the intensity of solar radiation is lower. Consequently, the solar still absorbs minimal heat,

and the evaporation of water is lesser. As the day progresses, the sun rises and reaches higher angle of incidence, leading to a higher concentrated solar intensity. The solar intensity increases, reaching a maximum (829.6 w/m²) around noon. The occurrence of maximum solar intensity around noon is similar to the studies of Osigbemeh et al [25]. During this period the efficiency of the solar still is higher following the availability of maximum thermal energy for evaporation. Post noon, the sun starts to descend, and hence the solar intensity declines. The decline in solar intensity gradually decreases the rate of evaporation in the afternoon, and hence the performance of the solar still reduces significantly. By late afternoon and into the evening, the solar intensity greatly reduces, and consequently, the evaporation rate. The fluctuation in solar intensity emphasizes the significance of modifications in the solar still design to harness solar energy effectively. Variation in Temperatures The effect of design modifications on the average absorber plate, glass plate, and water temperatures obtained across the three solar still configurations are illustrated in Figs. 5-7. The temperatures in all three solar still configurations rise from 10:00 hours and reach the maximum between 12:00 and 13:00 hours, and then a gradual decrease is witnessed in the afternoon. Higher temperatures during the midday impart more thermal energy to water, accelerating the evaporation of water [26]. In the conventional solar still (CSS), the maximum temperature of the absorber plate reaches around 68 °C when the solar radiation is peaking (Fig. 5). The maximum temperature supports the evaporation process, as the available heat is absorbed by the plate and the same is transferred to the water medium. Subsequently, the water temperature (Fig. 6) in the CSS lags behind the absorber plate temperature (66 °C). This phenomenon is consistent with the studies of Peng and Sharshir [27]. The glass plate, otherwise termed the condensation surface, maintains a lower

temperature (Fig. 7), around 55-57 °C, due to its exposure to the ambient air and the evaporative cooling effect [28]. The temperature gradient between the water and the glass plate is crucial for maximizing the condensation rate, thus improving the water production. However, the introduction of fins in a pyramid shaped solar still (PSSF) improves the overall heat transfer area, promoting better heat distribution to result in higher thermal performances than the CSS. The fins, attached to the absorber plate, facilitate more efficient heat distribution throughout the water [29]. In addition, the pyramid shape promotes concentration of solar radiation to amplify the absorber plate temperature. This modification raises the absorber plate temperature to 73-77 °C (Fig. 5). Consequently, the water temperature also rises, reaching 74 °C (Fig. 6), as the fins allow more heat to be absorbed and transferred. The glass plate temperature in this design reaches 60 °C (Fig. 7), but the increased evaporation rate leads to a higher overall productivity (detailed in the next section). The third solar still configuration, pyramid shaped solar still with fins covered by black cloth, displays the maximum

Figure 5. Variation in absorber plate temperature over time.

thermal performance. The usage of black cloth increases the absorptivity, minimizes the reflective losses, and maximizes the heat retention [30]. As a result, the absorber plate reaches a temperature as high as 75-78 °C (Fig. 5), higher than the other two attempted solar still configurations. Consequently, the water temperature increases to 75 °C (Fig. 6). The difference in temperature between the water and the glass plate, reaching nearly 64 °C (Fig. 7), significantly improves the evaporation and condensation rates. The black cloth-covered fins make sure that the heat is distributed uniformly and retained for a longer duration, leading to a higher temperature in the water. Singh et al [31] opined that enhancing the surface area enhances the absorber plate temperature, consistent with the present study. It is concluded that the variations in the absorber plate, water, and glass plate temperatures among the three configurations illustrate the profound influence of design modifications on solar still performance. Though the conventional pyramid shaped still is effective, the addition of fins and the use of black cloth over the fins provide significant improvements in attaining higher temperatures, leading to increased potable water production (discussed below).

Water Collected The conventional solar still (CSS) traps the solar energy through a transparent cover, which heats the water prevailing inside. Subsequently, the water evaporates and condenses on the cooler surface of the cover and is collected as distilled water. The CSS produces around 2.3 liters of water per day under optimal conditions (Fig. 8). However, the pyramid shaped solar still with fins (PSSF) yields 3 liters of water, 30% more than the CSS. The increased water yield is due to the higher surface area exposure to sunlight created by the geometrical modification i.e. the pyramid shaped absorber plate, as reported by Hammoodi et al [32]. The surface area and water evaporation rates of the solar still are further increased by the attachment of fins. The fins increase the temperature of the water and the inner surfaces of the solar still, yielding more water. The presence of fins aids in even distribution of heat across the solar still, suppressing the temperature gradients within the still. The fins guide the water towards the collection trough as well.

The maximum water collection (3.7 liters), which is 23 % more than the PSSF is obtained for the pyramid-shaped solar still with fins covered with black cloth (PSSFB) configuration. The 23% increase is attributed to the presence of black cloth which harnesses more solar radiation to obtain superior evaporation and collection efficiency. In addition, the capillary effect of drawing water upwards and spreading it evenly over a larger surface area supports the phenomenon. As the water vapor rises, it condenses on the transparent glasses of the still, resulting in a higher water production. The dual effect of improved thermal absorption and efficient water distribution makes the PSSFB an effective modification for achieving higher thermal performance and more water collection. Energy Balance In a solar still, energy balance accounts for heat transfer between the absorber, glass, and water through radiation, convection, and conduction processes. These mechanisms

regulate the thermal dynamics essential for efficient water evaporation and condensation. The convective heat transfer between water and glass cover [33]

Convective heat transfer from glass cover to atmosphere [33] (16)

Convective mode of heat transfer between basin and water [33] (18)

The yield of the solar still or the amount of water condensed on the inner surface of the glass

Similarly, heat transfer by radiation between water and glass cover [33]

Uncertainty Analysis The accuracy and precision of measurements are influenced by various factors, including the selection of instruments, calibration, testing conditions, observations, environmental factors, readings, and test design [34]. To ensure the tests are properly conducted, an uncertainty analysis is performed. Table 1 lists the instruments used in this study, along with their model, accuracy, and operating range. The smallest potential error for each instrument is calculated by dividing the smallest measurable reading, the minor meter reading, by the lowest possible output value. This calculation yields the instrument’s minimum possible error. The uncertainties of the measuring instruments employed in the study are determined using the equation provided in Ref. [34]. Based on the uncertainty analysis, the total uncertainty of the measuring instruments in this study is found to be within ±1.582%.

(10) The heat transfer between water and glass cover by evaporation [33] (11) (12) Radiation heat transfer between glass cover and atmosphere [33] (13) (14) (15)

Prediction By DNN Model

The DNN model was trained with the temperatures and water yield data collected during the experimental days. By varying the number of hidden layers and the quantity of neurons in the hidden layers, numerous models were constructed. The range of hyperparameters attempted and the optimal neurons in each hidden layer are shown in Table 2.

Table 1. Instruments employed along with accuracy and range S. No

The optimum neurons were determined by employing the Adam optimizer, as recommended by Saravanan et al [35]. The learning rates and decays of the Adam optimizer are presented in Table 3. A scatter plot comparing the actual and predicted temperatures, as well as the amount of water collected, provides valuable insights into the model’s accuracy (Figs. 9-12). In the scatter plot for both temperatures and water collected, a larger proportion of experimental data scatter around the diagonal line. The prevalence of data points concentrated in the closer proximity of the diagonal line indicates the existence of a strong correlation between DNN prediction and experimental outcomes. The strong correlation of predicted values by the DNN model indicates higher accuracy [36]. In this model, only a small number of experimental conditions deviate from the median, resulting in a high R² value of 0.96, demonstrating a strong goodness of fit. Table 4 presents the performance metrics for the DNN model’s accuracy. With a Mean Absolute Error (MAE) of 1.1986, a Mean Absolute Percentage Error (MAPE) of 0.8978, a Mean Square Error (MSE) of 0.0796, and a high R² value of 0.9601, the DNN model shows a strong alignment with the experimental results. Therefore, the deep neural network model is recommended for predicting temperatures and water collection in a solar still.

Figure 9. Comparison of absorber plate temperatures (experimental vs predicted).

Figure 10. Comparison of glass plate and water temperatures (experimental vs predicted).

Figure 11. Comparison of water temperatures (experimental vs predicted).

Figure 12. Comparison of water collected experimental vs predicted.

Nomenclature

The salient conclusions from this novel experimental study on the fabrication and testing of three solar still configurations are as follows:

1. The surface modifications in the pyramid shaped

absorber plate with fins improve the absorber plate, water, and glass temperatures.

2. The employment of fins augments the heat transfer area,

and covering the fins by a black cloth cover enhances heat retention and improves the evaporation rate.

3. The combined effect of the pyramid shape, fins, and

use of black cloth cover results in the highest water collection among the three attempted configurations, highlighting their capability to effectively address acute water shortages.

4. The water collected in the solar still having both fins

and the black cloth model is nearly twice that of the conventional solar still. The pyramid shaped solar still with fins attains 33% more efficiency than the conventional solar still. 5. The Deep Neural Network model is capable of predicting the temperature and water collected in a solar still with an accuracy of 96%.

Area of glass cover Absorber Plate area Conventional Solar Still Deep Neural Network Heat transfer coefficient due to convection (basin-water) hc(g-a) Convective heat transfer coefficient (glass cover-atmosphere) hc(w-g) Convective heat transfer coefficient (water-glass cover) he(w−g) Evaporative heat transfer coefficient (water-glass cover) hfg Latent heat of vaporization hr(g-a) Radiation heat transfer coefficient (glass cover and atmosphere) hr(w-g) Radiation heat transfer coefficient (glass cover and water) MAE Mean Absolute Error MAPE Mean Absolute Percentage Error me(w-g) Water vapour condensation rate pg Partial pressures of vapour at the glass surface PSSF Pyramid Solar Still with Fins PSSFB Pyramid Solar Still with Fins and Black Cloth pw Partial pressures of vapour at the water surface

qc(b-w) qc(g-a) qc(w-g) qe(w-g) qr(g-a) qr(w-g) R² ta tg tsky tw Yk yk Yk_mean εeffective εg εw σ

Convective mode of heat transfer (basin-water) Convective heat transfer (glass cover-atmosphere) Convective heat transfer (water-glass cover) Evaporative heat transfer (water-glass cover) Heat transfer by radiation (glass cover-atmosphere) Heat transfer by radiation (water and glass cover) Coefficient of determination Ambient temperature Temperature of glass cover Sky temperature Temperature of water Actual value Predicted value Mean actual value Effective emittance (glass cover-water surface) Upper glass emissivity Lower glass emissivity Stefan–Boltzmann constant

Data Availability Statement

The authors confirm that the data that supports the findings of this study are available within the article. Raw data that support the finding of this study are available from the corresponding author, upon reasonable request.

Conflict Of Interest

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Ethics

There are no ethical issues with the publication of this manuscript.

Statement On The Use Of Artificial Intelligence

Artificial intelligence was not used in the preparation of the article.

References

1. Therm Eng 2021;184:116233. [CrossRef]
2. Salhi H, Hadjira A, Jamil B. Statistical analysis of the solar diffuse fraction radiation using regression optimized artificial neural network model to predict analysis of longitudinal data in India. J Therm Eng thermal efficiency and water yield of tubular solar 2023;9:776-785. [CrossRef] still. Case Stud Therm Eng 2022;30:101750. [CrossRef]
3. Mehta P, Bhatt N, Bassan G, Said Z, ElCheikh A. Exploring stepped solar still developments with energy efficiency for a solar still fitted with ther- a case study for potable water provision in salt moelectric modules by ANFIS and PSO-enhanced farming regions. Sustain Energy Technol Assess neural network: a nanofluid application. Powder 2024;64:103700. [CrossRef] Technol 2021;385:185-198. [CrossRef] 1740 J Ther Eng, Vol. 11, No. 6, pp. 1729−1740, November, 2025
4. Shatar NM, Sabri MFM, Salleh MFM, Ani MH. Energy, exergy, economic, environmental analy- ZM, Essa FA. Enhancing rotating wick solar still sis for solar still using partially coated condensing performance with various porous breathable belt cover with thermoelectric cover cooling. J Clean designs and nanofluid. Case Stud Therm Eng Prod 2023;387:135833. [CrossRef] 2023;49:103205. [CrossRef]
5. Sahu SK, Natarajan SK, KAS, Suraparaju SK. Experimental investigation of a solar still combined Gopalsamy S, Nagaraj M, Almasoud N, et al. with phase change material (RT58) in southern Enhancing solar still thermal performance: the India climatic conditions. Proc Inst Mech Eng C J role of surface coating and thermal energy stor- Mech Eng Sci 2023;237:2635-2646. [CrossRef] age in repurposed soda cans. J Energy Storage
6. Elsheikh A, Hammoodi KA, Ibrahim A, Mourad 2024;77:109807. [CrossRef] AHI, Fujii M, Abd-Elaziem W. Augmentation and [30] Hammoodi KA, Dhahad HA, Alawee WH, Omara evaluation of solar still performance: a comprehen- ZM. A detailed review of the factors impacting sive review. Desalination 2023;555:117239. [CrossRef] pyramid type solar still performance. Alex Eng J
7. Alshqirate A, Awad AS, Al Alawin A, Essa MA. 2023;66:123–154. [CrossRef] Experimental investigation of solar still productiv- [31] Singh V, Kumar R, Saxena A, Dobriyal R, Tiwari S, ity enhancement of distilled water by using natural Singh DB. An analytical study on the effect of differ- fibers. Desalination 2023;553:116487. [CrossRef] ent photovoltaic technologies on enviro-economic
8. Abdullah AS, Panchal H, Alawee WH, Omara ZM. parameter and energy metrics of active solar desalt- Methods used to improve solar still performance with ing unit. Energy 2024;294:130851. [CrossRef] generated turbulence for water desalination – detailed
9. Hemmatian A, Kargarsharifabad H, Esfahlani review. Results Eng 2023;17:101251. [CrossRef] AA, Rahbar N, Shoeibi S. Improving solar still
10. Liu W, Wang Z, Liu X, Zeng N, Liu Y, Alsaadi FE. A sur- performance with heat pipe/pulsating heat pipe vey of deep neural network architectures and their appli- evacuated tube solar collectors and PCM: an exper- cations. Neurocomputing 2017;234:11-26. [CrossRef] imental and environmental analysis. Sol Energy
11. Abdou MA. Literature review: efficient deep neural 2024;269:112371. [CrossRef] networks techniques for medical image analysis. Neural Comput Appl 2022;34:5791-5812. [33] Venkatasamy M, Sivamani S, Venkatesan H,
12. Salhi H, Belkhiri L, Tiri A. Evaluation of diffuse Thangaraj M. Experimental study on productivity fraction and diffusion coefficient using statistical enhancement of an improvised inclined solar still. analysis. Appl Water Sci 2020;10:1-12. [CrossRef] Int J Ambient Energy 2018;39:165–175. [CrossRef]
13. Miroshnichenko L. Main characteristics of the sun. In: Solar-Terrestrial Relations: From Solar Activity Experimental and theoretical analysis of solar still to Heliobiology. Cham: Springer Int Publ; 2020. p. with solar pond for enhancing the performance of 7-20. [CrossRef] sea water desalination. Water Reuse 2023;13:620–
14. Osigbemeh MS, Asaolu OS, Nwachukwu AN. 633. [CrossRef] Sustainable solar power from wall mounted photo- [35] Saravanan S, Gajalakshmi K, Kumararaja K, voltaics. Energy Syst 2023;14:1007-1022. [CrossRef] Sivaraman B, Raghukandan K. Prediction of alu-
15. Rahul P, Ramanathan S, Dhivagar R, Karthi P, minium–stainless steel explosive clad strength Manivel M. A review on solar desalination tech- through machine learning. Sadhana 2023;48:140. niques using vacuum technology. Proc Inst Mech [CrossRef] Eng C J Mech Eng Sci 2023;237:3086–3102. [CrossRef] [36] Kumararaja K, Sivaraman B, Saravanan S.
16. Peng G, Sharshir SW. Progress and performance Performance evaluation of hybrid nanofluid-filled of multi-stage solar still – a review. Desalination cylindrical heat pipe by machine learning algo- 2023;565:116829. [CrossRef] rithms. J Therm Eng 2024;10:286-298.

Share and Cite