Deep-Neural-Network-Based Economic Model Predictive Control for Ultrasupercritical Power Plant
Jinghan Cui, Tianyou Chai, Xiangjie Liu
Abstract
The dynamic economic optimization of the ultrasupercritical (USC) boiler-turbine unit has become an important task in modern power plants. Economic model predictive control (EMPC) has recently developed to be a promising method for realizing the dynamic economy. This EMPC essentially requires a highly reliable model for USC dynamic prediction which could reflect the internal mechanism of USC with big data feature. This article constitutes a deep-neural-network-based EMPC for the USC unit. Deep belief network (DBN) is used to model the USC unit with mathematical structure. To overcome the nonlinearity and time delay existing in the pulverized channel, an augmented model with predictor embedded is also incorporated into the EMPC design. The auxiliary controller and stability region have been constituted to guarantee closed-loop stability. Simulation results on a 1000-MW USC unit fully demonstrate the effectiveness of the proposed DBN-based EMPC.