An established control strategy for biomass grate boilers based on a low-order nonlinear model is considered. Under ideal conditions, it achieves decoupled control of desired outputs by means of input–output linearization. The decoupling is gradually reduced and control performance deteriorates when actuator saturation occurs. This may be avoided by appropriately shaping the control strategy’s reference values. This contribution presents a method to do so by solving a sequence of linear programs. Its implementation requires the knowledge of typically unknown limits of mass-flows fed into the plant. An estimation strategy for these limits based on measurable quantities is thus proposed. Experimental data from three different scenarios is presented, in which the reference shaping improves tracking, mitigates wind-up phenomena and reduces emissions, respectively.