Thermal efficiency prediction of a solar low enthalpy steam generating plant employing artificial neural networks

Abstract

The present paper describes a mathematical model based on application of Artificial Neural Networks (ANN) employing a Multi- Layer Perceptron (MLP) model for thermal efficiency prediction of a solar low enthalpy steam generation plant composed by a Parabolic Trough Collector (PTCs) array. The MLP model uses physical data measurement in the steam prssoduction for heat processes. The input parameters used to achieve the prediction of thermal efficiency where: inlet and outlet working fluid temperature, flow working fluid, ambient temperature, direct solar radiation and wind velocity. After several training, the best MLP architecture was obtained employing Levenberg- Marquardt optimization algorithm, the logarithmic sigmoid transferfunction and the linear transfer-function for the hidden and output layer; and four neurons at the hidden layer, which predicts the thermal efficiency with a satisfactory determination coefficient (R2 = 0.99996). The predictive model can be implemented at intelligent sensors that allow to improve control of the PTCs system and leads to better utilization of the solar resource.