In thermal power systems, the quality of main steam temperature and main steam pressure control directly affects the safety and efficiency of thermal power units. In recent years, domestic power plants launched a variety of distributed control system (DCS) to achieve automatic control of main loops of power units. However, PID control algorithm built-in DCS behave not that good when controlling the main steam temperature and the main steam pressure, which are large time-delay, nonlinear and time-varying systems.As represent of predictive control algorithms, generalized predictive control (GPC) gained more applications in the field of industrial process control because of its output prediction, rolling optimization, adaptive ability, minimized model and etc. In recent years, the combination of generalized predictive control, multi-model strategies, experts systems and other advanced control strategies improved the control quality of thermal power unit, which is a large time-delay, strong nonlinear and time-varying system. These advanced methods are worth to be promoted.In this article, based on two thermal power units-#1and#3units in Shandong Shiheng power plant, a detailed study of the thermal power boiler process is made. Combined with Modbus, OPC communication technology, least squares method, stair-like generalized predictive control (SGPC) algorithm, expert system and multi-model switching strategies, a high portability, security and advanced control system is set up.The main findings are as follows:1)Single-level multi-model structure for the main steam temperature and multi-level multi-model structure for the main steam pressure are raised. Based on both, different controller and switching strategies are proposed.2) An expert supervision system is designed for handling soot-blowing process disturbance on the main steam temperature.3) A detailed study of the difficulties for the main steam temperature and main steam pressure control brought by the transformation of low-NOx combustion is made. After transformation of low NOx combustion, model mismatch and furnace combustion instability are serious. Initial models are modified. And pre-superheated temperature feed-forward control strategy is proposed to solve these problems. These are proved to be effective. |