Artificial Intelligence-Based Prediction of Modified Solar Still Productivity Using FGSVM and LOOCV

Dumka, Pankaj; Pandey, Nitesh; R. Mishra, Dhananjay; Chauhan, Rishika; Forootan Fard, Habib

Volume 2, Issue 2 (11-2026) AISEE 2026, 2(2): 91-99 | Back to browse issues page

Mendeley

Zotero

RefWorks

Dumka P, Pandey N, R. Mishra D, Chauhan R, Forootan Fard H. Artificial Intelligence-Based Prediction of Modified Solar Still Productivity Using FGSVM and LOOCV. , AISEE. 2026; 2 (2) :91-99
URL: http://aisesjournal.com/article-1-60-en.html

Artificial Intelligence-Based Prediction of Modified Solar Still Productivity Using FGSVM and LOOCV

Pankaj Dumka

, Nitesh Pandey

, Dhananjay R. Mishra

, Rishika Chauhan

, Habib Forootan Fard ^*

Department of Renewable Energies, Faculty of New Sciences and Technologies, University of Tehran, Tehran, Iran

Abstract: (10 Views)

Accurate prediction of freshwater productivity is key to design and optimise solar desalination systems. A Fuzzy Gaussian Support Vector Machine (FGSVM) model is developed in this paper to predict the hourly distillate yield of a sponge-augmented Modified Solar Still (MSS) under real climatic conditions. The experimental data were obtained from the single slope solar still integrated with sponges that enhance evaporation through increase in water retention, capillary action and effective use of thermal energy. Field experiments were conducted at Raghogarh, Guna, India (24°39′N, 77°19′E) using 40 kg of basin water mass. Input parameters were the basin water temperature (Tw), inner glass cover temperature (Tci), solar radiation intensity (I) and the temperature difference between basin water and glass cover (Tw - Tci). The hourly freshwater productivity was used as output variable. To ensure a robust model validation with a limited experimental dataset, training and testing were done using Leave-One-Out Cross Validation (LOOCV). The efficiency of the proposed FGSVM model was evaluated by Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and coefficient of determination (R2). The model gave an RMSE of 14.775, MAE of 11.224 and R2 of 0.980 which indicated an excellent agreement between observed and predicted productivity values. The results show that the FGSVM model can well capture the nonlinear thermal behaviour of the MSS, and the freshwater yield can be predicted accurately. Therefore, the proposed approach can be used as a reliable artificial intelligence tool for productivity forecasting, reducing experimental effort and supporting the development of intelligent and efficient solar desalination systems.

Keywords: Sponge-Augmented Solar Still, Fuzzy Gaussian Support Vector Machine, Artificial Intelligence, Solar Desalination, Machine Learning.

Full-Text [PDF 727 kb] (3 Downloads)

Type of Study: Research | Subject: General
Received: 2026/07/13 | Accepted: 2026/07/13 | Published: 2026/11/11

* Corresponding Author Address: India