Burning multi-fuel, including gases, liquid fuels and coal, whose flow rates and heating values vary all the time, a typical boiler in the steel and iron plant poses a challenge to achieving optimal operation. The present study proposes to develop an adaptive data-driven thermal efficiency estimator of multi-fuel boilers based on statistical identification of key variables. With the available on-line efficiency model, the model-based controller is hence readily applicable to improve the boiler efficiency. Real operation data taken from two industrial boilers are used to verify the effectiveness of the proposed method. The first half part of data serves to develop statistical models while the second half part serves to be simulated as virtual plants. The application of the proposed methods improved 1.94% of the thermal efficiency of a boiler burning multi-gas and 0.73% of a boiler burning coal and multi-gas in the virtual plant simulations. (C) 2009 Published by Elsevier Ltd.