Optimal operation policy of dynamic processes can be significantly different with and without due consideration to the process-model mismatches. In this work, a general optimisation framework is developed to obtain efficiently the optimal operation policy of dynamic processes under process-model mismatches. Neural network techniques have been used to predict the dynamic non-linear process-model mismatches. An important feature of the optimisation framework is that it allows the use of discrete process data in a continuous model to predict discrete and/or continuous mismatch profiles. The method is applied to an inherently dynamic batch distillation process and is demonstrated with an example. Neural network technique is found to predict dynamic mismatch profiles for this process with sufficient accuracy. Optimal operation:policy using a simple model with mismatches, predicted by the neural network, is found to be very close to that of the actual process.