This paper is concerned with using artificial neural networks for modelling and temperature control of a yeast fermentation biochemical reactor. At first, a neural model of the process is trained using available data sets generated from the fundamental model. The neural model is pruned in order to reduce its complexity and to improve its prediction ability. Next, a computationally efficient nonlinear model predictive control (MPC) algorithm with Nonlinear Prediction and Linearisation (MPC-NPL) which needs solving on-line a quadratic programming problem is developed. It is shown that the algorithm results in closed-loop control performance similar to that obtained in nonlinear MPC, which hinges on full on-line non-convex optimisation. The computational complexity of the MPC-NPL algorithm is shown, the control accuracy and the disturbance rejection ability are also demonstrated in the case of noisy measurements and disturbances affecting the process. (C) 2008 Elsevier B.V. All rights reserved.