Feedforward/state feedback laws are developed for on-line optimization of batch reactors in the presence of measurable disturbances. The degrees of singularity with respect to manipulated input and disturbance input are used to characterize the nature of the feedforward/state feedback laws. Explicit synthesis formulas are derived that relate the optimal manipulated input to the system states and the measurable disturbance. It is shown that when the degree of singularity with respect to the manipulated input is infinite, the end-point optimization problem reduces to a regulation problem in the presence of disturbances. For finite degrees of singularity, the optimal feedforward/state feedback law is, in general, dynamic. The proposed methodology is illustrated through several end-point optimization problems in batch reactors.