The paper documents an optimal multiple-model strategy to design a controller for the boiler-turbine unit. There are two important problems associated with multiple-model controllers: number of models and their locations in an operating space. A method based on Vinnicombe distance is used to find minimum number of local models. First the stability of the control scheme is verified in theory. Using the closed-loop stability condition the minimum number of models is found and the models are located in a space of an unknown parameter. The performance of the proposed controller is evaluated in simulation and compared with a PI controller designed using H-infinity approach and an MPC-based controller in the literature. The designed controller based on the linear local models can ensure satisfactory performance. The novel idea of this paper is the optimal strategy finding minimum number of local models in order to reduce the computational cost and processing time. (C) 2012 Elsevier Ltd. All rights reserved.