MGPC based on hopfield network and its application in a thermal power unit load system

Abstract

Multivariable General Predictive Control (MGPC) is an effective application in the control of plant with inertia and delay. But it has some defects such as requiring a large amount of computation online and poor treatment of constraints. This paper introduces Hopfield neural network into MGPC. Firstly, the MGPC was decomposed into several multi-input and single-output systems, then they were converted into several quadratic constrained optimizing problems. Several Hopfield networks were used to solve each quadratic constrained optimizing problem respectively. The Hopfield network has the merits of simple arithmetic and rapid computation. The combination of the two methods can overcome the defects of MGPC. Then the new method was applied to the control of a unit load system in a thermal power plant that is a 2×2 multivariable plant with coupling and constraints. Simulation proved that the new method has effective control performance. © Springer-Verlag Berlin Heidelberg 2006.