Utilization of LSTM neural network for water production forecasting of a stepped solar still with a corrugated absorber plate

Abstract

This study introduces a long short-term memory (LSTM) neural network model to forecast the freshwater yield of a stepped solar still and a conventional one. The stepped solar still was equiped by a copper corrugated absorber plate. The thermal performance of the stepped solar still is compared with that of conventional single slope solar still. The heat transfer coefficients of convection, evaporation, and radiation process have been evaluated. The exergy and energy efficiencies of both solar stills have been also evaluated. The yield of the stepped solar still is enhanced by about 128 % compared with that of conventional solar still. Then, the proposed LSTM neural network method is utilized to forecast the hourly yield of the investigated solar stills. Field experimental data was used to train and test the developed model. The freshwater yield was used in a time series form to train the proposed model. The forecasting accuracy of the proposed model was compared with those obtained by conventional autoregressive integrated moving average (ARIMA) and was evaluated using different statistical assessment measures. The coefficient of determination of the forecasted results has a high value of 0.97 and 0.99 for the conventional and the stepped solar still, respectively.