Applying extended automatic differentiation technique to process system optimization problems

An extended automatic differentiation (XAD) algorithm is developed to generate the derivative expressions which are coded into subroutines to replace the first derivative evaluation module of the SQP algorithm. Numerical results show that the derivative evaluation by derivative expressions, which is produced by XAD, is much faster than that by finite-difference approximation and plain automatic differentiation. A significant enhancement in optimization efficiency is also obtained by employing XAD technique instead of derivative approximation method