Membership function comparative investigation on productivity forecasting of solar still using adaptive neuro‐fuzzy inference system approach

Abstract

Modeling solar still productivity (SSP) is one of the most studied topics in solar desalination due to it having essential applications in the design of solar still systems. This study applied an adaptive neuro‐fuzzy inference system (ANFIS) and different membership functions (MFs) to predict the SSP required by designers, operators, and beneficiaries of solar stills. The output of this research can be used as a reference for designing and managing solar stills that could lead to optimizing the performance. The modeling process was based on real‐field experimental data. The model considers the solar radiation, relative humidity, total dissolved solids of the feed, total dissolved solids of the brine, and feed flow rate as the input variables. The results show that ANFIS forecasting with generalized bell MF (GBELLMF) produced the highest correlation coefficient (CC) and the smallest root mean square error (RMSE) when compared with other MF types. Thus, the ANFIS model with GBELLMF (CC = 0.99; RMSE = 0.03 L/m2/h) provides the best SSP prediction accuracy, which is better than other models with MFs. In addition, the statistical indicators demonstrate that the ANFIS model is better for predicting the SSP than multiple linear regressions. These findings demonstrate that ANFIS can be applied to forecast the SSP using weather and operational data as inputs with the best membership function (which is GBELLMF). © 2017 American Institute of Chemical Engineers Environ Prog, 37: 249–259, 2018