Solar thermal plants have high nonlinearities and non-manipulated energy source which make their control task a very challenging work. Linear controllers cannot cope with undesirable deviations of the outlet temperature over all the operation range of the dynamics of this type of plants. Moreover, nonlinear predictive control relying on online nonlinear optimization have the drawback of time consuming and numerical calculus issues. In this paper, an infinite gain scheduling neural predictive control is designed and applied to control the temperature in a distributed parabolic trough solar collector field. The performance of both tracking and disturbance rejection of the proposed controller is compared to those of four nonlinear predictive control variants: Two unconstrained neural predictive control using the Levenberg–Marquardt and the Broyden–Fletcher–Goldfarb–Shanno algorithms, and two constrained nonlinear predictive control using interior point algorithm, one is based on a neural network model and the other one is based on a first principal model. The superiority of the proposed control strategy is well demonstrated through simulation results.