Fault tolerant MPC of a solar trough field based on classification and regression trees

Direct solar radiation is important in a solar trough plant because it is the perturbation that affects most the operation. Generally, it is measured locally. Production losses or dangerous situations may occur when controlling a plant with wrong measurements. A possible solution is to have several pyrheliometers though two problems arise: cost and sensor fusion (complexity). The aim of this paper is to prove that a solar estimation based on Classification and Regression Trees can be used to design a Fault Tolerant Model Predictive Control strategy capable to work with erroneous values of radiation or even none, avoiding dangerous situations or production losses.