Energy production estimation of a parabolic trough solar power plant using artificial neural network

The aim of the present study is the use of Artificial Neural Networks (ANN) modeling to estimate the hourly based electric energy generation of a Parabolic Trough Solar Thermal Power Plant (PTSTPP), located in Eastern Morocco. Data covering 4 years are used in order to train and validate a three Multi-Layers Perceptron (MLP) model. In order to choose the best architecture, several statistical criteria are used such as: Coefficient of Correlation (R), Root Mean Squared Error (RMSE), Relative Root Mean Squared Error (RRMSE), Relative Mean Bias Error (RMBE) and Mean Absolute Error (MAE). The back propagation learning algorithm is used to train different ANN architectures. Predicted results indicates that the total electric energy accumulated for the validation year was about 42.6 GWh/year, representing an underestimation less than 5% from the recorded energy. The results indicate that the ANN model can successfully estimate the energy production of a solar power plant with parabolic trough collectors.