Application of Artificial Neural Network Models (MLP and RBF) and Support Vector Machine (SVM) to Estimate the Shadow in Flat-plate Solar Collectors in Iran

In this study, the amount of shadow in different types of flat-plate solar collectors according to the geographical conditions of Iran was estimated by using artificial neural network models (MLP and RBF) and Support Vector Machine (SVM). In this study, two types of LM and BR training algorithms with sigmoid tangent transfer function (TanSig) and different number of neurons in a hidden layer with k-fold cross validation method were used to create random datasets at each stage of modeling. The results showed that the MLP model with BR training algorithm and (5-23-1) structure, can create high-precision data similar to real values. The average MAPE and R2 statistics for the above model were estimated to be 0.42 ± 0.10 and 0.99±0.01, respectively. Also, there was no significant difference between the actual data and the predicted values (95% probability) at mean, variance and distribution. The results of sensitivity analysis showed that the distance of the absorber plate and the glass cover is the most important factor influencing the formation of shadows.