Application of neural networks to optimisation of process systems by dynamic programming

The possibility of employment of artificial neural networks for enhancement of dynamic programming in its application to the optimisation of process systems is studied in the paper. The neural networks are used in each subsystem to approximate the dependence of optimal value of subsystem's objective function on the variables describing the state of the process streams feeding the subsystem. The availability of the analytical form of the objective function greatly enhances the effectiveness of optimisation calculations in the both phases of dynamic programming application.