Predicting Pyramid Geometric Solar Water Still Plant Efficiency Using RBF Based Multi-layer Perceptron

Solar still plant helps to produce fresh water. Development of an accurate predictive model is required on order to monitor and maintain. The machine learning (ML) based model deploys time-series data analysis for the prediction of performance of the still water plant. A review of pyramid geometry water still plant is first prepared. That provides an understanding of components of the still. Recent efforts for improving solar water plant performance are reviewed. The model applies neural network, and is based on the improved weight initialized multi-layer perceptron (). The performance is measured using the experiments on water-production and predictive values of production. The results indicate the superiority of proposed IWMLP. RMSE is 0.0056. A comparison with MLP is also carried out which results in RMSE = 0.0153. The experimental results show that the use of (i) good quality of datasets for learning, (ii) normalization, and (iii) weight initialization strategies improve the prediction accuracy of the model algorithm.