A policy- and experience- driven neural network (PENN) and its application to SISO and MIMO process control

A policy- and experience- driven Neural Network (PENN) is proposed. Its applicability is confirmed by applying it to both a nonlinear SISO water-level control with and without time delay and a MIMO distillation control. By online learning of both the global policy and local experience, the PENN controller can follow the dynamic change of the processes. The training of network is simple and straightforward.