Generalized predictive control of ultra-supercritical power plant by using a composite weighted human learning optimization network

Abstract

The ultra-supercritical power plant is an advanced energy conversion system with high coal utilization and low emission. It is a complex system associated with high nonlinearity, strong coupling, and subjected to various uncertainties and interferences. This article introduces a nonlinear generalized predictive control based on a composite weighted human learning optimization network to incorporate both local linear modeling and local control into a multi-layer dynamic structure. Since composite weighted human learning optimization network can be considered as a linear-in-weight network, the limitation of the non-convex is elaborately broken through incorporating composite weighted human learning optimization models into the convex quadratic programming routine of local linear generalized predictive control. Detailed analysis is given to show the validity and advantage of the proposed method in controlling the ultra-supercritical power plant.