Potential Assessment of Modelling Performance of a Central Tower Receiver Power Plant Using Intelligent Control Technique

Abstract

Recently, central tower receiver (CTR) power plant has been received a considerable attention as a promising technology for large solar thermal plants compared to different concentrated solar power (CSP) technologies. The present work adopted a facile controllable scheme using an artificial neural network (ANN) technique for modelling and simulating CTR plant with thermal energy storage (TES). Three different ANN models such as radial basis function (RBF), generalized regression neural network (GRNN), and multi-layer perceptron (MLP) were applied to assess the performance of CTR plant model. Based on statistical error analysis, MLP model was the optimal model compared to RBF and GRNN models. It is found that MLP model displays the best values for the coefficient of determination (R2=1), root mean square error (RMSE=0.003) and mean absolute error (MAE=0.0023) during ANN testing process. While, the values of R2, RMSE, and MAE were 0.999, 0.4817, and 0.32, respectively for GRNN model. Similarly, for RBF model, the values of R2, RMSE, and MAE were 0.9985, 0.2846, and 0.0674, respectively. The MLP provides a precise control over the discharge rate of the heat transfer fluid (HTF). Therefore, the receiver outlet temperature remains constant at the desired value regardless of the variations in direct solar radiation and receiver inlet temperature. Also, in this work, the algorithm of electrical generation methodology was modified for regulating CTR/TES output according to the hot storage tank (HST) conditions. The adopted model performance for a 40 MWe CTR power plant is validated by comparing its results with the obtained simulation results by System Advisor Model (SAM) software. The simulation results exhibit that the adopted CTR-ANN model and SAM results are in good agreement with each other. The reasonable simplicity and minimum required input data of CTR-ANN model make it an adequate tool to predict and analyse the performance of CTR technology in a simple and fixable manner.