Using artificial neural networks to predict the performance of a liquid metal reflux solar receiver: Preliminary results
Three and four-layer backpropagation artificial neural networks have been used to predict the power output of a liquid metal reflux solar receiver. The networks were trained using on-sun test data recorded at Sandia National Laboratories in Albuquerque, New Mexico. The preliminary results presented in this paper are a comparison of how different size networks train on this particular data. The results give encouragement that it will be possible to predict output power of a liquid metal receiver under a variety of operating conditions using artificial neural networks.