Optimizing Neural Net based Predictive Control

Abstract

Neural networks hold great promise for application in the general area of process control. This paper focuses on using a backpropagation network in an optimization based model predictive control scheme. Since analytical expressions for the gradient and Hessian of the neural net model can be derived and these expressions can be calculated in paralle, extremely fast computation times are possible. The control approach is illustrated on a pH CSTR example.