For real-time optimization of batch processes

a batch-to-batch self-optimizing control approach is proposed based on necessary conditions of optimality approximation method. Firstly

the nominal optimal input trajectory is numerically obtained

which is further parameterized as a few decision variables to simplify the problem complexity. Secondly

the optimality conditions for the batch-to-batch optimization problem are analyzed

from which regression models are built using terminal measurements. The obtained regression models are further adopted as controlled variables to be tracked from batch to batch. The proposed method is applied to a simulated batch reactor. The results show that the new method can realize batch-wise self-optimizing control efficiently.