Comparison between physical modelling and neural network modelling of a solar power plant

The paper presents results from the first phase of a research project carried out at the solar power plant located at Plataforma Solar de Almeria (PSA), Spain. The final goal of this project is to develop a control scheme in order to optimise the thermal energy extracted from the plant using a model-based predictive control strategy. In this contribution, the modelling and identification of the nonlinear plant is presented in two different approaches. The first approach is the development of a neural network model, while the second approach is the development of a physical model. The non-linear models are validated with data collected during experiments performed at PSA. Further on. the two models are compared in order to decide which one gives better results in approximating the behaviour of the plant.