Urea selective catalytic reduction (urea-SCR) systems are well known for exhaust gas aftertreatment in power plant applications. Stringent specifications and the dynamic operation mode of such applications demand advanced control strategies. The control purpose of urea-SCR is to simultaneously achieve high NOx conversion efficiency and low ammonia slip. The two emission demands are contradictory. Moreover, owing to the dynamic operation mode of urea-SCR systems, advanced control strategies are required to improve urea injection control. A data-driven predictive controller for urea injection was designed. The input-output data were directly obtained from ETC driving cycle test of the diesel engine bench. The controller was derived by the coupling excitation and separation of the data. Because of the physical characteristics, the input and output constraints were considered explicitly in the problem formulation. In order to obtain offset-free control for the reference input, the predictor equation was gained with incremental inputs and outputs. The experimental test was carried out in ETC and other transient test cycle. The results showed that the controller was able to meet emission control requirements under the excitation test cycle. The closed-loop system had the ability to reject uncertainties caused by uncertainty of working conditions under the non-excitation test cycle.