Modelling the free response of a solar plant for predictive control

Abstract

This paper demonstrates the identification of a nonlinear plant using neural networks for predictive control. The problem of neural identification is tackled using a static (non-recurrent) neural network in an autoregressive configuration (NARX). The selection of a set of input variables, a set of input/output vectors for training, and a neural structure, is discussed. In particular, an algorithm is proposed to obtain the number of past values of the measured variables needed to feed the network. The neural model is then used within a model-based predictive control (MBPC) framework. The MBPC scheme uses the prediction of the output of the system calculated as the sum of the free response (obtained using the nonlinear neural model) and the forced response (obtained linearizing around the current operating point) to optimize a performance index. The on-line adaptation of the model and other issues are discussed. The control scheme has been applied and tested in a solar power plant.