An Algorithm for Gradient-based Dynamic Optimization of UV Flash Processes
Flash processes are relevant for production optimization of oil reservoirs where the temporal evolution of the reservoir temperature is significant and when compositional models are used.
We consider optimal control problems in which differential-algebraic constraints represent dynamic vapor-liquid equilibrium processes without spatial effects. The dynamic optimization method single shooting is used to transcribe the optimal control problem into a nonlinear program (NLP) and an adjoint algorithm is developed for computing gradients that are necessary for a gradient-based optimization algorithm to solve such an NLP.
We consider an example of optimal tracking for a dynamic UV (isochoric-isoenergetic) flash process and compare two variations of numerical integration of the differential-algebraic equations.