Multi-input-multi-output (MIMO) controllers designed for the megawatt output/throttle pressure control in a boiler/turbine unit of a coal-fired power plant using the H ∞ optimization have been recently shown to exhibit performance and robustness significantly superior to that of the currently employed nonlinear PID-based controller. However, to fully realize the potential of robust boiler/turbine control, a sufficiently simple plant model, referred to as the nominal model, needs to be found together with an appropriate uncertainty description suitable for the subsequent robust controller synthesis. These two components form a conceptual paradigm, supported by a group of identification methods, that is referred to as identification for robust control. In the present work, several identification methods for robust control are discussed and their advantages and disadvantages compared. The main two of these methods - the identification in H ∞ and the stochastic embedding are found, however, to be not well suited for application to a boiler/turbine unit of a coal-fired power plant due to the information, unobtainable a priori, required for the former and the inability of the latter to handle data of the necessary length. To address this problem, a simple sinusoidal sweep method is combined with the identified nominal plant model, yielding a plant uncertainty model readily incorporable into the robust control design. Simulation results show that the methodology proposed provides an adequate solution to the problem of system identification for robust control of a boiler/turbine unit.