Neural network inverse models of supercritical boiler unit for intelligent coordinated controller design

A supercritical (SC) or ultra-supercritical (USC) once-through boiler unit is a typical multi-variable system with large inertia and non-linear, slow time-variant, time-delay characteristics, which often makes the coordinated control quality deteriorate under wide-range load-changing conditions, and thus influences the unit load response speed and leads to heavy fluctuations for main steam pressure. To improve the supercritical boiler unit's coordinated control quality with advanced intelligent control strategy, the neural-network based inverse system models for a 600MW supercritical boiler unit were investigated. The inputs and outputs of the two separate models for load and main steam pressure were determined by analyzing the schematic of the supercritical power unit and its coordinated control modes. A standard BP neural network and a BP neural network with time-delay inputs and time-delay outputs feedback were respectively adopted to establish the inverse models. The models were compared and validated by simulation tests, which showed that the models with time-delay inputs and outputs feedback are favorable for intelligent coordinated controllers' design with higher precision, better generalization ability and also simple structure.